Audit report automatic generation method based on natural language processing

Abstract

The invention discloses an audit report automatic generation method based on natural language processing, and relates to the field of data processing. The auditing report automatic generation method based on natural language processing comprises the steps of obtaining a plurality of auditing file sets corresponding to a target auditing stage, wherein the auditing file sets comprise a plurality of auditing files in the same time interval; according to the audit file frequency of each audit file set, screening to obtain a first audit file corresponding to a core audit event period in a target audit stage; according to the association influence degree of each first audit file on the core audit event period, determining the first audit file of which the association influence degree is greater than or equal to a preset influence degree threshold as a second audit file; and filling auditing data in each second auditing file into a corresponding position in a preset auditing template to obtain a target auditing report corresponding to the target auditing stage.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to a method for automatically generating an audit report based on natural language processing. Background Art

[0002] Audit reports are usually intended to provide a fair and independent review of the audited object to ensure that it fairly reflects the relevant conditions and operating results in all material aspects. The generation of traditional audit reports is through the auditors' layer-by-layer analysis and review of audit materials, which involves a lot of data analysis and document writing, and the whole process takes a long time. In order to ensure the completeness and accuracy of the report, it will also go through multiple reviews and rechecks during the period. Therefore, if you want to obtain a high-quality audit report as soon as possible, it is very demanding on the auditors' experience and judgment.

[0003] However, the existing methods are affected by the subjective judgment of auditors, and the final audit report has low accuracy. In addition, due to the complexity and diversity of audit data, the efficiency of manually generating audit reports is low. Summary of the invention

[0004] The embodiment of the present invention provides a method for automatically generating an audit report based on natural language processing, which can improve the accuracy and generation efficiency of the audit report.

[0005] In one aspect of an embodiment of the present invention, a method for automatically generating an audit report based on natural language processing is provided, comprising: Acquire multiple audit file sets corresponding to the target audit phase, where the audit file sets include multiple audit files in the same time period; According to the audit file frequency of each audit file set, the first audit file corresponding to the core audit event period in the target audit stage is screened; According to the correlation influence of each first audit file on the core audit event period, the first audit file whose correlation influence is greater than or equal to the preset influence threshold is determined as the second audit file; The audit data in each second audit file is filled into the corresponding position in the preset audit template to obtain the target audit report corresponding to the target audit stage.

[0006] In the method for automatically generating an audit report based on natural language processing provided by an embodiment of the present invention, a preset audit template is provided. By analyzing the audit files in the multiple audit file sets corresponding to the target audit stage, the corresponding audit data is automatically extracted and filled in according to the preset audit template. In this way, the audit report can be automatically generated without the need to manually generate the audit report, thereby improving the efficiency of generating the audit report. At the same time, according to the audit file frequency of the multiple audit file sets corresponding to the target audit stage, the first audit file corresponding to the core audit event period in the target audit stage is obtained. Based on this, the correlation influence of each first audit file on the core audit event period is further analyzed, and the second audit file that can have an important impact on the core audit event period is selected from each first audit file. In this way, the key second audit file is obtained through comprehensive analysis, and the audit report is generated according to the second audit file, so that the comprehensive audit data performance can be reflected and the accuracy of the audit report is improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0007] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings required for use in the embodiments or the prior art descriptions are 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.

[0008] Figure 1 A flowchart of a first method for automatically generating an audit report based on natural language processing provided by an embodiment of the present invention; Figure 2 A flowchart of a second method for automatically generating an audit report based on natural language processing provided by an embodiment of the present invention; Figure 3 A flowchart of a third method for automatically generating an audit report based on natural language processing provided by an embodiment of the present invention; Figure 4 A flowchart of a fourth method for automatically generating an audit report based on natural language processing provided by an embodiment of the present invention; Figure 5 A flowchart of a fifth method for automatically generating an audit report based on natural language processing provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0009] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the following is a detailed description of the specific implementation method, structure, features and effects of a method for automatically generating an audit report based on natural language processing proposed by the present invention in combination with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.

[0010] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0011] It should be noted that the acquisition, storage, use, and processing of data in the technical solution of the present invention are in compliance with the relevant provisions of laws and regulations.

[0012] It should be noted that in the embodiments of the present invention, certain software, components, models and other existing solutions in the industry may be mentioned, which should be regarded as exemplary. Their purpose is only to illustrate the feasibility of implementing the technical solution of the present invention, but it does not mean that the applicant has or will necessarily use the solution.

[0013] In the existing methods, auditors usually manually analyze and organize data, and then manually edit the corresponding audit report. However, this method is affected by the subjective judgment of auditors, and the final audit report is less accurate. And due to the complexity and diversity of audit data, the efficiency of manually generating audit reports is low.

[0014] The purpose of the present invention is to provide a method for automatically generating an audit report based on natural language processing. In the method for automatically generating an audit report based on natural language processing provided by an embodiment of the present invention, a preset audit template is provided. By analyzing the audit files in the multiple audit file sets corresponding to the target audit stage, the corresponding audit data is automatically extracted and filled in according to the preset audit template. In this way, the audit report can be automatically generated without the need for manual generation of the audit report, thereby improving the efficiency of generating the audit report. At the same time, according to the audit file frequency of the multiple audit file sets corresponding to the target audit stage, the first audit file corresponding to the core audit event period in the target audit stage is obtained. Based on this, the correlation influence of each first audit file on the core audit event period is further analyzed, and the second audit file that can have an important impact on the core audit event period is selected from each first audit file. In this way, the key second audit file is obtained through comprehensive analysis, and the audit report is generated according to the second audit file, so that the comprehensive audit data performance can be reflected and the accuracy of the audit report is improved.

[0015] Figure 1A flow chart of a method for automatically generating an audit report based on natural language processing is provided. The method for automatically generating an audit report based on natural language processing can be applied to a server. The method for automatically generating an audit report based on natural language processing can include the following S101 to S104.

[0016] S101, obtaining a plurality of audit file sets corresponding to a target audit phase, wherein the audit file sets include a plurality of audit files in a same time interval.

[0017] In this embodiment, the target audit period is a time interval pre-selected for auditing. For example, the target audit period may be the mth month.

[0018] The audit files in the audit file set belong to the same time period. For example, the audit files can be divided into units of each day, and the audit files on the same day are divided into the same audit file set. The audit files can include the company's audit report, purchase order, project file, contract file, etc.

[0019] As an example, the server first exports all audit files in the mth month from the enterprise audit system, and divides each audit file into units of each day to obtain multiple audit file sets corresponding to the mth month.

[0020] S102, screening out the first audit file corresponding to the core audit event period in the target audit phase according to the audit file frequency of each audit file set.

[0021] In this embodiment, a large number of audit files are generated during the core audit event period, and they often last for a fixed period of time. Therefore, the core audit event period can be screened out based on the magnitude relationship of the audit file frequencies of the audit file set.

[0022] The audit file frequency of an audit file set is the number of audit files included in the audit file set.

[0023] Specifically, the target audit phase can be divided according to the preset division days to obtain multiple time phases. Then, for each time phase, the sum of the corresponding audit file frequencies is calculated, and the time phase corresponding to the maximum sum of audit file frequencies is determined as the core audit event period; in the embodiment of the present invention, the preset division days are set to 1, which can be adjusted according to the specific implementation environment.

[0024] S103: According to the correlation influence of each first audit file on the core audit event period, a first audit file whose correlation influence is greater than or equal to a preset influence threshold is determined as a second audit file.

[0025] In this embodiment, the correlation influence is used to characterize the correlation between the first audit file and the core audit event period. The greater the correlation between the first audit file and the core audit event period, the greater the correlation influence; the smaller the correlation between the first audit file and the core audit event period, the smaller the correlation influence.

[0026] As an example, the server performs statistics on each first audit file to obtain the top 10 keywords with the highest frequency in each first audit file to form a keyword set. Then, for each first audit file, similarity calculation is performed with the keyword set to obtain the correlation influence corresponding to each first audit file. Finally, each first audit file whose correlation influence is greater than or equal to the preset influence threshold is determined as the second audit file.

[0027] S104, filling the audit data in each second audit file into the corresponding position in the preset audit template to obtain a target audit report corresponding to the target audit stage.

[0028] In this embodiment, the preset audit template is a template of a preset audit report. For example, the audit report template may include a transaction product introduction area, a transaction amount introduction area, and a customer qualification introduction area.

[0029] As an example, the server first extracts the audit data from each second audit file, and then fills each audit data into the corresponding area in the preset audit template, so as to obtain the target audit report corresponding to the target audit stage.

[0030] In the method for automatically generating an audit report based on natural language processing provided in this embodiment, a preset audit template is provided. By analyzing the audit files in the multiple audit file sets corresponding to the target audit stage, the corresponding audit data is automatically extracted and filled in according to the preset audit template. In this way, the audit report can be automatically generated without the need for manual generation of the audit report, thereby improving the efficiency of generating the audit report. At the same time, according to the audit file frequency of the multiple audit file sets corresponding to the target audit stage, the first audit file corresponding to the core audit event period in the target audit stage is obtained. Based on this, the correlation influence of each first audit file on the core audit event period is further analyzed, and the second audit file that can have an important impact on the core audit event period is selected from each first audit file. In this way, the key second audit file is obtained through comprehensive analysis, and the audit report is generated according to the second audit file, so that the comprehensive audit data performance can be reflected and the accuracy of the audit report is improved.

[0031] As an optional embodiment, Figure 2 As shown, S102 may specifically include the following S201-S204: S201, sorting the audit file frequencies of each audit file set according to the time sequence of the audit file set to obtain an audit file frequency sequence; S202, splitting the target audit phase into multiple event periods according to the extreme value points in the audit file frequency sequence; S203, obtaining a core audit event period among multiple event periods according to the audit file frequencies of each audit file set corresponding to each event period; S204: Determine the audit file in each audit file set corresponding to the core audit event period as the first audit file.

[0032] In this embodiment, extreme value points include maximum value points and minimum value points. Maximum value points are the maximum points corresponding to each rising trend, and minimum value points are the minimum points corresponding to each falling trend. Extreme value points represent specific audit events or market changes, so extreme value points can be used as boundaries to divide event periods.

[0033] As an example, after obtaining the audit file set corresponding to each day of the mth month, the server counts the number of audit files included in each audit file set, and sorts the number of audit files in chronological order to obtain an audit file frequency sequence.

[0034] Then, the server further finds the extreme points in the audit file frequency sequence, and divides it into multiple event periods with the extreme points as the boundary. Among them, the extreme points can be preset to belong to the previous event period or the next event period. For example, suppose the audit file frequency sequence is {6, 8, 10, 15, 11, 7, 5, 9, 14}. Among them, the extreme points are 15, 5 and 14, so the event periods are divided with these three points as the boundary. It can be obtained that 6, 8, 10, 15 belong to one event period, 11, 7, 5 belong to one event period, and 9, 14 belong to one event period.

[0035] Then, the audit file frequency of each audit file set in each event period is averaged, that is, the average daily audit file frequency in each event period is obtained. The event period with the largest average daily audit file frequency is determined as the core audit event period, and the audit file in each audit file set corresponding to the core audit event period is determined as the first audit file.

[0036] Through this embodiment, the target audit phase is divided into multiple event periods with specific audit significance, and the core audit event period and its corresponding first audit file are accurately identified. This can help to accurately locate the audit files that need to be focused on during the audit process, thereby ensuring a comprehensive review and accurate evaluation of key audit information and improving the accuracy of the audit report.

[0037] As an optional embodiment, S203 may specifically include: Perform mean processing on the audit file frequencies of the audit file set in each event period to obtain the mean value of the audit file frequency in each event period; The difference between the audit file frequencies of the audit file sets in adjacent time periods in each event period is processed by averaging to obtain the floating mean of the frequency in each event period; Divide the frequency mean of each audit file by the corresponding frequency floating mean to obtain the concentrated appearance of files in each event period; The event period corresponding to the largest appearance in the file set is determined as the core audit event period.

[0038] In this embodiment, the server calculates the mean of the audit file frequency of the audit file set in each event period, thereby obtaining the mean of the audit file frequency corresponding to each event period. Then, the difference of the audit file frequency of the two adjacent days before and after each event period is obtained, and the mean of the difference is calculated, thereby obtaining the floating mean of the frequency corresponding to each event period.

[0039] Then, according to the frequency mean of each audit file and the corresponding frequency floating mean, the occurrence degree of the file concentration in each event period is determined, and the event period with the largest occurrence degree of the file concentration is determined as the core audit event period.

[0040] Specifically, the occurrence of files in the event period can be determined by the following formula 1: Formula 1 In the formula, Tb represents the occurrence degree of the file set in the b-th event period, represents the mean frequency of audit files during the b-th event period, When the floating mean frequency of the b-th event period is 0, the denominator in Formula 1 is replaced by the constant 0.1 to avoid the situation where the denominator is 0.

[0041] Among them, the average frequency of audit files in the bth event period is The larger the value is, the more files appear in the b-th event period, and the higher the corresponding file concentration is; the floating mean of the frequency in the b-th event period is The larger it is, the greater the floating degree of the file in the bth event period, and the smaller the corresponding occurrence degree in the file set.

[0042] The larger the appearance degree Tb of the files in the bth event period is, the more likely it is that a large number of audit files appeared in the bth event period and remained stable for a period of time, thus making it more likely that the bth event period is the core audit event period.

[0043] Through this embodiment, a comprehensive analysis is conducted from two aspects: the mean frequency of audit files in each event period and the mean frequency fluctuation, to obtain the possibility that each event period belongs to the core audit event period, and the event period with the greatest possibility is determined as the core audit event period. In this way, the core audit event period can be accurately determined through the mean frequency of audit files and the mean frequency fluctuation, which helps to accurately obtain the second audit file from the core audit event period in the future, and improves the accuracy of the audit report.

[0044] As an optional embodiment, Figure 3 As shown, S103 may specifically include the following S301-S304: S301, obtaining a keyword sequence corresponding to each first audit file; S302, according to the occurrence frequency of the keywords in each keyword sequence, combining the keywords whose occurrence frequency meets the preset frequency condition to obtain a high-frequency keyword sequence; S303, comparing each keyword sequence with the high-frequency keyword sequence respectively, and determining the correlation influence of each first audit file on the core audit event period; S304: Determine the first audit file whose correlation impact is greater than or equal to a preset impact threshold as the second audit file.

[0045] In this embodiment, the preset frequency condition is a pre-set condition for filtering keywords to obtain high-frequency keywords. For example, the preset frequency condition can be keywords ranked in the top 10% of frequency or keywords that appear more than 10 times.

[0046] The preset impact threshold is a minimum condition of the second audit file that is highly relevant to the core audit event period and is preset according to business requirements. For example, the preset impact threshold may be 0.8.

[0047] As an example, the server first uses term frequency-inverse document frequency (TF-IDF) or a text mining algorithm to extract keywords from each first audit file to form a keyword sequence corresponding to each first audit file.

[0048] Then, the server counts the frequency of occurrence of each keyword in all keyword sequences according to each keyword sequence, and obtains the high-frequency keyword sequence by ranking the keywords in the top one-third in frequency.

[0049] Then, a comparison algorithm is used to compare the similarity between the keyword sequence of each first audit file and the high-frequency keyword sequence, and the similarity is used as the correlation impact score. The comparison algorithm can use indicators such as cosine similarity or Jaccard similarity.

[0050] Finally, the first audit documents with an associated influence degree greater than or equal to 0.6 are determined as the second audit documents. These documents are considered to be highly relevant to the core audit event period and require key review and analysis.

[0051] Through this embodiment, according to the keyword sequences corresponding to each first audit document, the associated influence degree of each first audit document on the core audit event period is calculated. Thus, the second audit documents highly relevant to the core audit event period can be effectively identified and processed, thereby improving the efficiency and accuracy of the audit work.

[0052] As an alternative embodiment, S301 may specifically include: Perform word segmentation on the first audit document to obtain a first word segmentation sequence corresponding to the title of the first audit document and a second word segmentation sequence corresponding to the content of the first audit document; Calculate the similarity between each title word segmentation in the first word segmentation sequence and the second word segmentation sequence to obtain the first importance degree of each title word segmentation; Combine the title word segmentations with a first importance degree greater than or equal to the preset importance degree threshold to obtain the keyword sequence corresponding to the first audit document.

[0053] In this embodiment, the preset importance degree threshold is the minimum condition that the first importance degree of the word segmentation belonging to the keyword needs to meet, which is set in advance. By way of example, the preset importance degree threshold may be 0.88.

[0054] As an example, for the first audit document, the server may first perform word segmentation on the title and content of the first audit document through the jieba word segmentation technology, so as to obtain multiple title word segmentations corresponding to the title of the first audit document and multiple content word segmentations corresponding to the content of the first audit document.

[0055] On this basis, the server removes the common pause words and irrelevant words therein, such as "de", "le", etc., so as to obtain a first word segmentation sequence corresponding to the title of the first audit document and a second word segmentation sequence corresponding to the content of the first audit document.

[0056] Then, the server calculates the similarity between each title word segmentation in the first word segmentation sequence and the second word segmentation sequence to obtain the first importance degree of each title word segmentation. Specifically, the first importance degree of the title word segmentation can be determined by the following formula 2: Formula 2 In the formula, represents the first importance degree of the z-th title word segmentation, jac() represents the jaccard coefficient, z represents the z-th title word segmentation, and B represents the second word segmentation sequence corresponding to the content of the first audit document.

[0057] Among them, That is, to calculate the correlation between the zth topic word and the second topic word sequence. The higher the correlation, the more important the topic word is, and the more it can express the core idea of ​​the first audit document, that is, the higher the first importance at this time.

[0058] Finally, it is determined whether the first importance of the topic segmentation words is greater than 0.88, and the topic segmentation words whose first importance is greater than or equal to 0.88 are combined to obtain a keyword sequence corresponding to the first audit file.

[0059] Through this embodiment, the similarity between the first word segmentation sequence corresponding to the title of the first audit file and the second word segmentation sequence corresponding to the content of the first audit file is calculated, so that the keywords that can express the core idea of ​​the first audit file can be accurately identified, which is helpful for subsequently calculating the correlation influence of the first audit file on the core audit event period based on the keyword sequence of the first audit file, and can improve the efficiency and accuracy of the audit work.

[0060] As an optional embodiment, Figure 4 As shown, S303 may specifically include the following S401-S404: S401, comparing the keyword sequence with the high-frequency keyword sequence to obtain the high-frequency keyword corresponding to the high-frequency keyword sequence included in the keyword sequence; S402, performing mean processing on the occurrence frequency of each high-frequency keyword to obtain an average frequency of the high-frequency keywords in the keyword sequence; S403, calculating the similarity between the keyword sequence and the high-frequency keyword sequence to obtain the second importance of the keyword sequence; S404: Determine the correlation influence of the first audit file corresponding to the keyword sequence on the core audit event period by using the average frequency of the high-frequency keywords and the second importance.

[0061] In this embodiment, the server first divides the number of occurrences of each high-frequency keyword in the high-frequency keyword sequence by the total number of occurrences of the high-frequency keywords in each first audit file to obtain the frequency of occurrence of each high-frequency keyword.

[0062] Then, the server compares the extracted keyword sequence with the high-frequency keyword sequence to find multiple high-frequency keywords included in the keyword sequence, and calculates the average of the occurrence frequencies of the multiple high-frequency keywords to obtain the average frequency of the high-frequency keywords.

[0063] Then, the similarity between the keyword sequence and the high-frequency keyword sequence is calculated according to the above formula 2 to obtain the second importance of the keyword sequence.

[0064] Finally, the average frequency of high-frequency keywords and the second importance are combined to determine the correlation influence of the first audit file corresponding to the keyword sequence on the core audit event period through a comprehensive model (such as weighted sum, product, etc.). For example, a formula can be set to multiply the average frequency of high-frequency keywords by the corresponding first weight factor, and the second importance by the corresponding second weight factor, and finally add them up to get the correlation influence of the first audit file on the core audit event period.

[0065] Through this embodiment, based on the average frequency of high-frequency keywords corresponding to the keyword sequence and the second importance of the keyword sequence, the correlation influence of the first audit file on the core audit event period can be accurately calculated, thereby improving the efficiency and accuracy of the audit work.

[0066] As an optional embodiment, S404 may specifically include: Multiply the average frequency of high-frequency keywords by the second importance to get the initial influence; The initial impact is normalized and calculated to obtain the correlation impact of the first audit file corresponding to the keyword sequence on the core audit event period.

[0067] In this embodiment, the correlation influence of the first audit file on the core audit event period can be determined by the following formula 3: Formula 3 In the formula, represents the correlation influence of the first audit file in the mth month on the core audit event period, jac() represents the jaccard coefficient, G1 represents the high-frequency keyword sequence in the core audit event period, The keyword sequence representing the first audit file of the nth month, It represents the average frequency of high-frequency keywords in the first audit file of the mth month and the nth audit file. norm represents normalized calculation.

[0068] in, It indicates the second importance of the keyword sequence. The larger the second importance, the more important the first audit document is, the more it can express the core idea of ​​the core audit event period, and the greater the correlation influence on the core audit event period; the larger the average frequency of high-frequency keywords, the more important the high-frequency words appearing in the first audit document are, and the greater the correlation influence on the core audit event period.

[0069] The correlation influence of the first audit file in the mth month on the core audit event period The larger the value is, the greater the correlation between the nth first audit file and the core audit event period of the mth month, that is, the greater the possibility of the second audit file belonging to the mth month.

[0070] Through this embodiment, based on the average frequency of high-frequency keywords corresponding to the keyword sequence and the second importance of the keyword sequence, the correlation influence of the first audit file on the core audit event period can be accurately calculated, thereby improving the efficiency and accuracy of the audit work.

[0071] As an optional embodiment, Figure 5 As shown, S104 may specifically include the following S501-S503: S501, using the first segmentation sequence and the second segmentation sequence respectively corresponding to each second audit file, determining the correlation distance between each second audit file; S502, clustering the second audit files according to the association distances between the second audit files to obtain at least one clustered file cluster; S503, filling the audit data in each cluster file cluster into the position corresponding to the cluster file cluster in the preset audit template, and obtaining a target audit report corresponding to the target audit stage.

[0072] In this embodiment, the clustering algorithm for clustering the second audit file may be K-means clustering, DBSCAN clustering or hierarchical clustering.

[0073] As an example, the server first calculates the association distance between the second audit files according to the first word segmentation sequence and the second word segmentation sequence of the second audit file by using cosine similarity, Euclidean distance or Jaccard similarity.

[0074] Then, the server uses the K-means clustering algorithm to cluster the target number of clustered file clusters according to the correlation distance between each second audit file. The server can obtain multiple historical audit reports, calculate the average of the number of report directories in each historical audit report, and use the average as the target number. Alternatively, the server can obtain the number of report directories of a preset audit template and use the number of report directories of the preset audit template as the target number.

[0075] Finally, the audit data in each cluster file cluster is extracted, such as abnormal transactions and non-compliant behaviors, and the audit data of each cluster file cluster is filled into the position corresponding to the cluster file cluster in the preset audit template, so as to obtain the target audit report corresponding to the target audit stage. Specifically, the similarity between each cluster file cluster and each report directory in the preset audit template can be calculated using cosine similarity, Euclidean distance or Jaccard similarity. And the audit data of each cluster file cluster is filled into the report directory with the highest similarity in the preset audit template.

[0076] Through this embodiment, each second audit file is clustered into at least one cluster file cluster according to the correlation distance between each second audit file. Finally, the audit data in each cluster file cluster is filled into the corresponding position in the preset audit template to generate a target audit report. In this way, a large number of audit files can be efficiently processed and analyzed, and similar audit contents can be automatically identified and classified, so as to quickly generate accurate and comprehensive audit reports, which can improve the efficiency and effectiveness of audit work.

[0077] As an optional embodiment, S501 may specifically include: Calculate the similarity between the first word segmentation sequence corresponding to the first file and the first word segmentation sequence corresponding to the second file to obtain a first distance between the first file and the second file, where the first file and the second file are any two files that are different from each other in the second audit files; Calculate the similarity between the second word segmentation sequence corresponding to the first file and the second word segmentation sequence corresponding to the second file to obtain a second distance between the first file and the second file; The first distance is multiplied by the second distance to obtain the association distance between the first file and the second file.

[0078] In this embodiment, the association distance between the second audit files can be determined specifically by the following formula 4: Formula 4 In the formula, represents the association distance between the first file and the second file, jac() represents the jaccard coefficient, represents the first word sequence of the first file, represents the first word segmentation sequence of the second file, Represents the second word segmentation sequence of the first file, The second word segmentation sequence represents the second document.

[0079] in, represents the first distance between the first file and the second file. The higher the similarity between the first word segmentation sequence of the first file and the first word segmentation sequence of the second file, the greater the association distance between the two. It represents the second distance between the first file and the second file. The higher the similarity between the second word segmentation sequence of the first file and the second word segmentation sequence of the second file, the greater the association distance between the two.

[0080] The correlation distance between the first file and the second file The larger the value is, the more similar the first file is to the second file, and the greater the possibility that they belong to the same cluster file cluster.

[0081] Through this embodiment, the association distance between the first file and the second file can be accurately determined by considering the first distance of the first word segmentation sequence and the second distance of the second word segmentation sequence between the first file and the second file. Therefore, accurate clustering can be performed according to the association distance, thereby improving the reliability of the clustering result.

[0082] As an optional embodiment, before S503, the method for automatically generating an audit report based on natural language processing may further include: Obtain multiple historical audit reports prior to the target audit phase; According to the report catalogues of each historical audit report, the intersection of the report catalogues of each historical audit report is obtained; Generate a preset audit template based on the intersection of the report catalog.

[0083] In this embodiment, the server obtains multiple historical audit reports in a preset historical stage before the target audit stage, and then obtains the report directory of each historical audit report.

[0084] Then, use the VLOOKUP function or INDEX / MATCH function to calculate the intersection between the report directories, and design a new preset audit template based on the intersection between the report directories.

[0085] Through this embodiment, by effectively utilizing the information of historical audit reports, a comprehensive and accurate preset audit template is generated for a new audit project, thereby facilitating the subsequent direct generation of an audit report corresponding to the target audit stage based on the preset audit template, thereby improving the efficiency of generating audit reports.

[0086] It should be clear that the present invention is not limited to the specific configuration and processing described above and shown in the figures. For the sake of simplicity, a detailed description of the known method is omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of the present invention is not limited to the specific steps described and shown, and those skilled in the art can make various changes, modifications and additions, or change the order between the steps after understanding the spirit of the present invention.

[0087] It should also be noted that the exemplary embodiments mentioned in the present invention describe some methods or systems based on a series of steps or devices. However, the present invention is not limited to the order of the above steps, that is, the steps can be performed in the order mentioned in the embodiments, or in a different order from the embodiments, or several steps can be performed simultaneously.

[0088] The above is only a specific implementation of the present invention. Those skilled in the art can clearly understand that for the convenience and simplicity of description, the specific working process of the system, module and unit described above can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here. It should be understood that the protection scope of the present invention is not limited to this. Any technician familiar with the technical field can easily think of various equivalent modifications or replacements within the technical scope disclosed by the present invention, and these modifications or replacements should be covered within the protection scope of the present invention.

Claims

1. A method for automatically generating an audit report based on natural language processing, characterized in that: The method comprises: Acquire multiple audit file sets corresponding to the target audit phase, wherein the audit file sets include multiple audit files in the same time interval; According to the audit file frequencies of each of the audit file sets, a first audit file corresponding to the core audit event period in the target audit phase is obtained by screening; According to the correlation influence of each of the first audit files on the core audit event period, determining the first audit files whose correlation influence is greater than or equal to a preset influence threshold as second audit files; The audit data in each of the second audit files is filled into the corresponding position in the preset audit template to obtain the target audit report corresponding to the target audit stage.

2. The method for automatically generating an audit report based on natural language processing according to claim 1 is characterized in that: The step of screening and obtaining the first audit file corresponding to the core audit event period in the target audit phase according to the audit file frequency of each audit file set includes: Sorting the audit file frequencies of each of the audit file sets according to the chronological order of the audit file sets to obtain an audit file frequency sequence; According to the extreme value points in the frequency sequence of the audit file, the target audit phase is divided into multiple event periods; Obtaining a core audit event period among the multiple event periods according to the audit file frequencies of each audit file set corresponding to each event period; The audit files in each of the audit file sets corresponding to the core audit event period are determined as the first audit files.

3. The method for automatically generating an audit report based on natural language processing according to claim 2 is characterized in that: The obtaining of the core audit event period in the plurality of event periods according to the audit file frequencies of the audit file sets corresponding to the event periods comprises: Performing mean processing on the audit file frequencies of the audit file set in each of the event periods to obtain the mean of the audit file frequencies in each of the event periods; Performing mean processing on the differences between the audit file frequencies of the audit file set in adjacent time periods in each of the event periods to obtain a floating mean of the frequencies in each of the event periods; Divide the frequency mean of each audit file by the corresponding frequency floating mean to obtain the concentrated appearance of files in each event period; The event period corresponding to the largest occurrence degree in the file set is determined as the core audit event period.

4. The method for automatically generating an audit report based on natural language processing according to claim 1 is characterized in that: The step of determining, according to the correlation influence of each of the first audit files on the core audit event period, the first audit files having the correlation influence greater than or equal to a preset influence threshold as second audit files comprises: Obtaining a keyword sequence corresponding to each of the first audit files; According to the occurrence frequency of the keywords in each of the keyword sequences, the keywords whose occurrence frequency meets the preset frequency condition are combined to obtain a high-frequency keyword sequence; Compare each of the keyword sequences with the high-frequency keyword sequence to determine the correlation influence of each of the first audit files on the core audit event period; The first audit file whose association impact is greater than or equal to the preset impact threshold is determined as the second audit file.

5. The method for automatically generating an audit report based on natural language processing according to claim 4 is characterized in that: The obtaining of the keyword sequence corresponding to each of the first audit files includes: Performing word segmentation processing on the first audit file to obtain a first word segmentation sequence corresponding to the title of the first audit file and a second word segmentation sequence corresponding to the content of the first audit file; Calculate the similarity between each topic segmentation in the first segmentation sequence and the second segmentation sequence to obtain the first importance of each topic segmentation; The topic segments whose first importance is greater than or equal to a preset importance threshold are combined to obtain a keyword sequence corresponding to the first audit file.

6. The method for automatically generating an audit report based on natural language processing according to claim 4 is characterized in that: The step of comparing each of the keyword sequences with the high-frequency keyword sequence to determine the correlation influence of each of the first audit files on the core audit event period includes: Compare the keyword sequence with the high-frequency keyword sequence to obtain the high-frequency keywords corresponding to the high-frequency keyword sequence included in the keyword sequence; Performing mean processing on the occurrence frequency of each of the high-frequency keywords to obtain an average frequency of the high-frequency keywords of the keyword sequence; Calculate the similarity between the keyword sequence and the high-frequency keyword sequence to obtain the second importance of the keyword sequence; The average frequency of the high-frequency keywords and the second importance are used to determine the correlation influence of the first audit file corresponding to the keyword sequence on the core audit event period.

7. The method for automatically generating an audit report based on natural language processing according to claim 6 is characterized in that: The determining the correlation influence of the first audit file corresponding to the keyword sequence on the core audit event period by using the average frequency of the high-frequency keywords and the second importance includes: Multiplying the average frequency of the high-frequency keywords by the second importance to obtain an initial influence; The initial impact is normalized and calculated to obtain the correlation impact of the first audit file corresponding to the keyword sequence on the core audit event period.

8. The method for automatically generating an audit report based on natural language processing according to claim 1 is characterized in that: The step of filling the audit data in each of the second audit files into the corresponding position in the preset audit template to obtain the target audit report corresponding to the target audit stage includes: Determine the correlation distance between each of the second audit files by using the first segmentation sequence and the second segmentation sequence respectively corresponding to each of the second audit files; Clustering the second audit files according to the association distances between the second audit files to obtain at least one cluster file cluster; The audit data in each of the clustered file clusters is filled into the position corresponding to the clustered file cluster in the preset audit template to obtain the target audit report corresponding to the target audit stage.

9. The method for automatically generating an audit report based on natural language processing according to claim 8 is characterized in that: Determining the association distance between the second audit files by using the first word segmentation sequence and the second word segmentation sequence respectively corresponding to the second audit files includes: Calculate the similarity between the first word segmentation sequence corresponding to the first file and the first word segmentation sequence corresponding to the second file to obtain a first distance between the first file and the second file, wherein the first file and the second file are any two files that are different from each other in the second audit files; Calculate the similarity between the second word segmentation sequence corresponding to the first file and the second word segmentation sequence corresponding to the second file to obtain a second distance between the first file and the second file; The first distance is multiplied by the second distance to obtain an association distance between the first file and the second file.

10. The method for automatically generating an audit report based on natural language processing according to claim 8, characterized in that: Before filling the audit data in each of the clustered file clusters into a position corresponding to the clustered file cluster in a preset audit template to obtain a target audit report corresponding to the target audit stage, the method further includes: Obtain multiple historical audit reports before the target audit stage; According to the report catalogues of the historical audit reports, an intersection of the report catalogues of the historical audit reports is obtained; The preset audit template is generated according to the intersection of the report directories.

Images