Financial statement recognition method, device, computer equipment and readable storage medium

Abstract

The present invention discloses a method, device, computer equipment and readable storage medium for identifying financial statements, the method comprising: receiving a pending financial report picture of a pending financial report; identifying the pending financial report picture to obtain a financial report document contained in the pending financial report picture, the financial report document comprising layout information, subject information and financial report data information; inputting the layout information and the subject information into a subject matching model, outputting a matching result, the matching result comprising the matching subject of the pending financial report and the confidence level of the matching subject; determining the target subject corresponding to the pending financial report from the matching subject according to the confidence level, and generating a target financial report based on the target subject and the financial report data information. Thereby, the recognition accuracy of financial statements is improved, and it is convenient for users to maintain financial statements. In addition, the present invention also relates to blockchain technology, and the target financial report can be stored in the blockchain.

Description

Technical Field

[0001] The embodiments of the present invention relate to the field of financial statements, and in particular to a method, device, computer equipment and readable storage medium for identifying financial statements. Background Art

[0002] Financial statements are accounting statements that reflect the capital and profit status of an enterprise or budget unit over a certain period of time. The types, formats, and reporting requirements of my country's financial statements are all stipulated by a unified accounting system, requiring enterprises to compile and report them regularly. At present, state-owned industrial enterprises should separately report capital balance sheets, special funds and special appropriation tables, capital construction loans and special loan tables, and other capital statements, as well as profit statements, product sales profit details, and other profit statements at the end of the reporting period; state-owned commercial enterprises must submit capital balance sheets, operating status statements, and special funds statements. Financial statements include balance sheets, profit and loss statements, cash flow statements or financial position changes statements, schedules, and notes.

[0003] The importance of financial reports to financial leasing companies is self-evident, and it runs through the entire life cycle of financial leasing. Common financial report platforms on the market only manage financial report templates in a unified manner, and further perform OCR recognition on paper financial reports. Due to the problem of existing OCR recognition accuracy, there are still a lot of operations that need to be reviewed by human eyes after recognition, which is inefficient and prone to errors. Summary of the invention

[0004] In view of this, an object of an embodiment of the present invention is to provide a method, apparatus, computer device and readable storage medium for identifying financial statements, which improves the recognition accuracy and facilitates users to perform financial statement maintenance.

[0005] To achieve the above object, an embodiment of the present invention provides a method for identifying a financial statement, comprising:

[0006] Receive the pending financial report pictures of the pending financial report;

[0007] Identify the financial report image to be processed to obtain a financial report document contained in the financial report image to be processed, wherein the financial report document includes format information, subject information, and financial report data information;

[0008] Inputting the format information and the subject information into a subject matching model, and outputting a matching result, wherein the matching result includes a matching subject of the to-be-processed financial report and a confidence level of the matching subject;

[0009] According to the confidence level, a target subject corresponding to the to-be-processed financial report is determined from the matching subjects, and a target financial report is generated based on the target subject and the financial report data information.

[0010] Furthermore, the identification of the to-be-processed financial report image to obtain a financial report document contained in the to-be-processed financial report image, wherein the financial report document includes layout information, subject information and financial report data information including:

[0011] Preprocessing the financial report to be processed to obtain a standard image;

[0012] Performing layout recognition on the standard image to obtain a corresponding financial report layout, wherein the layout information includes the financial report layout and the layout name;

[0013] Performing character recognition on the standard image to obtain multiple field information, the field information including the format name, subject information and financial report data information;

[0014] The layout of the financial report is restored according to the field information, and a verification is performed to obtain an identified financial report document.

[0015] Furthermore, the format information and the subject information are input into a subject matching model, and a matching result is output, wherein the matching result includes the matching subject of the to-be-processed financial report and the confidence level of the matching subject, including:

[0016] Inputting the format information and the subject information into a subject matching model, so as to obtain a subject template library matching the format information according to the format information through the subject matching model;

[0017] Segmenting the subject information according to the granularity of characters to obtain character information corresponding to each subject, wherein the subject information includes a plurality of subjects;

[0018] Matching the character information corresponding to each of the subjects according to a preset inverted index table to obtain a subject candidate set corresponding to each subject from the subject template library;

[0019] Each of the subjects is matched with a standard subject in a subject candidate set corresponding to the subject, and a matching result is output, wherein the matching result includes a matching subject of the financial report to be processed and a confidence level of the matching subject.

[0020] Furthermore, before matching the character information corresponding to each subject according to the preset inverted index to obtain the subject candidate set corresponding to each subject, the method includes:

[0021] Get multiple standard subjects;

[0022] A plurality of inverted index tables are established by using the first characters of the plurality of standard subjects as keywords, and the plurality of inverted index tables are stored in the subject template library.

[0023] Furthermore, each of the subjects is matched with a standard subject in a subject candidate set corresponding to the subject, and a matching result is outputted. The matching result includes the matching subject of the to-be-processed financial report and the confidence level of the matching subject, including:

[0024] Calculate the similarity value between each subject and the standard subject in the subject candidate set corresponding to the subject according to the similarity algorithm;

[0025] The standard subjects whose similarity values ​​are greater than a preset threshold are taken as matching subjects, and the confidence levels of the matching subjects are determined according to the similarity values ​​of the matching subjects.

[0026] Furthermore, the method further comprises:

[0027] Get multiple financial report data;

[0028] Performing data analysis on the target financial report and the financial report data to obtain analysis results;

[0029] Generate warning information based on the analysis results.

[0030] Furthermore, the method further comprises:

[0031] The target financial report is stored in the blockchain.

[0032] To achieve the above purpose, an embodiment of the present invention further provides a financial statement recognition device, comprising:

[0033] A receiving module, used for receiving the financial report picture to be processed of the financial report to be processed;

[0034] An identification module, used for identifying the financial report image to be processed to obtain a financial report document contained in the financial report image to be processed, wherein the financial report document includes format information, subject information and financial report data information;

[0035] A matching module, used for inputting the format information and the subject information into a subject matching model, and outputting a matching result, wherein the matching result includes the matching subject of the to-be-processed financial report and the confidence level of the matching subject;

[0036] A generation module is determined, which is used to determine the target subject corresponding to the to-be-processed financial report from the matching subjects according to the confidence level, and to generate a target financial report based on the target subject and the financial report data information.

[0037] To achieve the above-mentioned purpose, an embodiment of the present invention further provides a computer device, which includes a memory and a processor, wherein the memory stores a computer program that can be run on the processor, and when the computer program is executed by the processor, the steps of the financial statement identification method as described above are implemented.

[0038] To achieve the above objectives, an embodiment of the present invention further provides a computer-readable storage medium, in which a computer program is stored. The computer program can be executed by at least one processor so that the at least one processor performs the steps of the financial statement identification method as described above.

[0039] The financial statement recognition method, device, computer equipment and readable storage medium provided in the embodiments of the present invention obtain layout information, subject information and financial report data information by recognizing the financial report image to be processed, and then input the layout information, subject information and financial report data information into the subject matching model to perform subject matching through the subject matching model, thereby further correcting the subject, improving the recognition accuracy, and facilitating users to perform financial report maintenance. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 The present invention is a flowchart of a first embodiment of a method for identifying financial statements.

[0041] Figure 2 This is a flowchart of step S120 in the first embodiment of the present invention.

[0042] Figure 3 This is a flowchart of step S140 in the first embodiment of the present invention.

[0043] Figure 4 This is a flowchart of step S147 in the first embodiment of the present invention.

[0044] Figure 5 This is a flowchart of step S144 in the first embodiment of the present invention.

[0045] Figure 6 This is a flowchart of step S170 in the first embodiment of the present invention.

[0046] Figure 7 This is a schematic diagram of program modules of Embodiment 2 of the financial statement recognition device of the present invention.

[0047] Figure 8 This is a schematic diagram of the hardware structure of the computer device embodiment 3 of the present invention. DETAILED DESCRIPTION

[0048] In order to make the purpose, technical scheme and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0049] Embodiment 1

[0050] See also Figure 1 , showing a flowchart of the steps of the method for identifying financial statements in the first embodiment of the present invention. It can be understood that the flowchart in the embodiment of the method is not used to limit the order of executing the steps. The following is an exemplary description using the computer device 2 as the execution subject. The details are as follows.

[0051] Step S100, receiving the financial report picture to be processed.

[0052] Specifically, the financial report to be processed is a financial statement that needs to be identified, and the image of the financial report to be processed may be a scanned copy or a photo of the financial report to be processed. The photo of the financial report to be processed in the image of the financial report to be processed is relatively straight and neat.

[0053] Step S120, identifying the financial report image to be processed to obtain the financial report document contained in the financial report image to be processed, wherein the financial report document includes layout information, subject information and financial report data information.

[0054] Specifically, the financial report to be processed is identified through OCR technology, multiple field information is identified, and the multiple field information is analyzed to obtain the financial report type, financial report name, subject, and financial report data. The obtained financial report type can identify the layout of the financial report to be processed, and then match it with the built-in financial report template library. Different types of financial report layouts, or standard financial report templates are different, so the corresponding financial report type is obtained; the financial report contains type information, and after obtaining multiple field information through text recognition, it is matched with the fields of various standard financial report types to determine the financial report type; it can also be obtained through manual identification by the user. Subject information and financial report data information can be identified through key values, and a correlation is established. The corresponding financial report name is identified to obtain the main title, and the layout information includes the financial report layout, financial report type, and financial report name.

[0055] For example, see Figure 2 , the step S120 comprises:

[0056] Step S121, pre-processing the financial report to be processed to obtain a standard image.

[0057] Specifically, image input: For different image formats, there are different storage formats and different compression methods. Preprocessing: Mainly includes binarization, noise removal, tilt correction, etc. Binarization: Most of the pictures taken by the camera are color images. Color images contain a huge amount of information. The content of the picture can be simply divided into foreground and background. In order to make the computer recognize text faster and better, it is necessary to process the color image first so that the image only has foreground information and background information. The foreground information can be simply defined as black and the background information as white. This is the binarization image. Noise removal: For different documents, the definition of noise can be different. Denoising according to the characteristics of noise is called noise removal. Tilt correction: Since ordinary users are more casual when taking pictures of documents, the pictures taken will inevitably be tilted, which requires text recognition software to correct. After the above preprocessing, a standard picture is obtained.

[0058] Step S122, performing layout recognition on the standard image to obtain a corresponding financial report layout, wherein the layout information includes the financial report layout and the layout name.

[0059] Specifically, the process of dividing the document image into paragraphs and lines is called layout analysis. After the layout analysis, you can match it through the built-in templates. Different types of financial report templates are different, and get the financial report layout, such as general enterprises.

[0060] Step S123, performing character recognition on the standard image to obtain a plurality of field information, wherein the field information includes the format name, subject information and financial report data information.

[0061] Specifically, character recognition can be performed through template matching and feature extraction. Due to the influence of factors such as text displacement, stroke thickness, broken strokes, adhesion, and rotation, the difficulty of feature extraction is greatly affected. Character recognition is performed on standard images through a text recognition model. Due to the limitations of shooting conditions, characters often stick together and break strokes, which greatly limits the performance of the recognition system. This requires the text recognition software to have a character cutting function. The extracted features can be recognized by a classifier, which classifies them and outputs which text the feature corresponds to. The spectral clustering algorithm, K nearest neighbor algorithm, and K value parameter space automatic search algorithm are used to determine the recognition of the segmented text and identify which text it is. After recognition, the layout name, subject information, and financial report data information are obtained, and the recognition results are corrected according to the relationship between the specific language context.

[0062] Step S124, restoring the layout of the financial report according to the field information, and verifying it to obtain an identified financial report document.

[0063] Specifically, the recognized text is still arranged like the original document image, with the paragraphs, positions and order unchanged, and is output to a word document, pdf document, etc. This process is called layout recovery. After the recovery is completed, a financial report document is obtained. The fields of the account can be filled, and then the financial report data information can be directly filled into the target template.

[0064] Exemplarily, when the identified field content is not in the system financial report template library, the template can be automatically refined and supplemented.

[0065] Step S140, inputting the format information and the subject information into a subject matching model, and outputting a matching result, wherein the matching result includes the matching subjects of the to-be-processed financial report and the confidence level of the matching subjects.

[0066] Specifically, the layout information includes the financial report layout and the financial report name. The field information of the financial report layout, financial report name and subject information is input into the subject matching model as input data; after receiving the three parameters passed in, the subject matching model obtains the subject template set of the financial report template corresponding to the layout information, performs similarity calculations on the corresponding subjects, and generates subject matching results; the subject matching model outputs the matching results in the form of a dictionary, and the output content includes the matched subjects, confidence levels, and subjects to be selected. The financial report template is the financial report layout obtained after OCR recognition and analysis, such as general enterprises; the financial report name is the name of the financial report to be processed, such as cash flow statement, balance sheet, income statement, etc.; the subject is the subject after OCR technology recognition, such as notes receivable, etc. This step is to perform a rough recognition of financial report subjects, that is, to organize the text recognized by OCR.

[0067] For example, see Figure 3 , the step S140 comprises:

[0068] Step S141: input the format information and the subject information into a subject matching model, so as to obtain a subject template library matching the format information according to the format information through the subject matching model.

[0069] Specifically, the layout information includes the financial report layout and the financial report type. The field information of the financial report template, financial report name and account information is input into the account matching model as input data. After the account matching model receives the three passed in parameters, it obtains the account template set of the financial report template corresponding to the layout information. The account template set includes multiple standard accounts.

[0070] Step S143, segmenting the subject information according to the granularity of characters to obtain character information corresponding to each subject, wherein the subject information includes multiple subjects.

[0071] Specifically, each subject in the subject information is segmented into at least two characters.

[0072] Step S145: Match the character information corresponding to each subject according to a preset inverted index table to obtain a candidate set of subjects corresponding to each subject from the subject template library.

[0073] Specifically, the character information of each subject is matched with the subject template library according to the inverted index table. In the inverted index table, the standard characters of standard subjects obtained from the standard templates are prestored, and each standard subject is numbered. As long as there is a match between the standard characters corresponding to a certain standard template and the characters of the subject, the standard subject corresponding to the standard character is recalled. Character similarity algorithms such as the longest common substring, edit distance algorithm, and cosine similarity algorithm in NLP technology can be used for similarity matching. For example: When using the edit distance algorithm, calculate how many times the characters of the subject need to be edited to obtain the standard characters of the standard subject. The number of modifications is the similarity. The fewer the number of times, the more similar it is. The edit operations include replacing one character with another, inserting a character, and deleting a character. Another example: When using the longest common substring, find the length of the same string in the characters of the subject and the standard characters of the standard subject in sequence. The longer the length, the greater the similarity. Then find the number corresponding to the standard subject in the inverted index table and associate it with the matching subject. The template corresponding to the number can be queried in the inverted index table in the order of the numbers to see if it contains the character information of the subject. If it contains the subject, recall the standard subject corresponding to the number from the subject template library to obtain the candidate set of subjects.

[0074] Step S147: Match each subject with the standard subjects in the candidate set of subjects corresponding to the subject, and output the matching result. The matching result includes the matching subject of the financial report to be processed and the confidence level of the matching subject.

[0075] Specifically, calculate the similarity between each subject and the standard subject corresponding to each subject according to the similarity algorithm, and determine the confidence level according to the similarity value. The matching result includes the matching subject, the confidence level of the matching subject, and the candidate subject. The matching subject is the financial report subject corresponding to the subject identified by the OCR technology; Confidence level: The accuracy of the match, such as exact match, algorithm match, fuzzy match, manual intervention, etc.; Candidate subject: When an exact match and an algorithm match cannot be performed, the candidate matching subject is given for output and extracted during manual intervention.

[0076] Example of matching subject: {

[0077] 'A': ['Jia', 1, []];

[0078] 'B': ['B', 2, []];

[0079] 'C': ['C', 3, ['big','small', 'many', 'few']];

[0080] 'D': ['D', 4, ['big','small','many', 'few']]

[0081] }

[0082] Note: A, B, C, and D are financial report subjects recognized by OCR. A, B, C, and D represent the recognized financial report subjects, and Jia, Yi, Bing, and Ding represent the matching subjects. 1, 2, 3, and 4 represent the recognition confidence levels. When the recognition is inaccurate (3, 4), Bing and Ding represent the candidate subjects with the highest similarity. The content in ['big','small','many', 'few'] represents the candidate financial report subjects to be manually confirmed.

[0083] Exemplarily, refer to Figure 4 , step S147 includes:

[0084] Step S147A, calculating the similarity value between each of the subjects and the standard subject in the subject candidate set corresponding to the subject according to the similarity algorithm.

[0085] Specifically, the similarity algorithm calculates the similarity value between each subject and the standard subject in the subject candidate set corresponding to the subject, and is not limited to using algorithms such as Euclidean distance, Pearson correlation coefficient, or cosine similarity to calculate the similarity value.

[0086] Step S147B, using the standard subject with the similarity value greater than the preset threshold as the matching subject, and determining the confidence level of the matching subject according to the similarity value of the matching subject.

[0087] Specifically, a preset threshold of the similarity value is set in advance, and the similarity values greater than the preset threshold are divided into confidence levels, and the confidence level is determined by the similarity value.

[0088] Exemplarily, refer to Figure 5 , before step S145, it includes step S144:

[0089] Step S144A, obtaining a plurality of standard subjects.

[0090] Specifically, according to the standard financial report type (enterprise type), obtain the standard financial report template of each enterprise, and extract all the subjects of a single standard template

[0091] Step S144B, creating a plurality of inverted index tables with the first character of the plurality of standard subjects as keywords, and storing the plurality of inverted index tables in the subject template library.

[0092] Specifically, all subjects of the standard template are indexed by using NLP and search engine technology at the granularity of words of the standard subject, and an inverted index table of each standard template can be generated in turn.

[0093] Step S160, determining the target subject corresponding to the to-be-processed financial report from the matching subjects according to the confidence level, and generating a target financial report based on the target subject and the financial report data information.

[0094] Specifically, if the confidence level is high or medium-high, corresponding to 1 and 2 in the above example, the matching standard subject is used as the target subject; if the confidence level is low or medium-low, corresponding to 3 and 4 in the above example, the matching standard subject is used as the candidate subject, indicating that the subject and the standard subject do not completely match, and the recommended matching standard subject is given according to the confidence of the subject and the standard subject, and then several fuzzy matching standard subjects (such as: ['big', 'small', 'more', 'less']) are obtained, and then manually verified to select the target subject. After the target subject is selected, the target subject is re-filled into the financial report document to obtain the target financial report.

[0095] For example, see Figure 6 , the method further comprises step S170:

[0096] Step S171, obtaining multiple financial report data.

[0097] Specifically, the financial report data of other listed companies in the same industry, that is, other financial report data, are pulled.

[0098] Step S172, performing data analysis on the target financial report and the financial report data to obtain analysis results.

[0099] Specifically, the target financial report is compared with the financial report data, and customized financial report analysis rules and warning trigger conditions are supported based on the industry characteristics of each company, the current economic environment, the company's operating data and other dimensions. At each stage of a financial leasing transaction with a company, including before, during and after the lease, the company's financial report data will be automatically analyzed in real time to obtain analysis results.

[0100] Step S173: Generate warning information based on the analysis result.

[0101] Specifically, early warning signals are given based on the results of data analysis to guide business personnel to conduct risk identification and quota assessment in a timely manner, and to conduct intelligent analysis and early warning of the company's financial statements throughout the entire life cycle of financial leasing.

[0102] Exemplarily, the method further includes:

[0103] The target financial report is stored in the blockchain.

[0104] Specifically, uploading the target financial report to the blockchain can ensure its security and fairness and transparency to users. The user device can download the summary information from the blockchain to verify whether the target financial report has been tampered with. The blockchain referred to in this example is a new application model of computer technologies such as distributed data storage, point-to-point transmission, consensus mechanism, and encryption algorithm. Blockchain is essentially a decentralized database. It is a string of data blocks generated by cryptographic methods. Each data block contains a batch of network transaction information, which is used to verify the validity of its information (anti-counterfeiting) and generate the next block. The blockchain can include the blockchain underlying platform, the platform product service layer, and the application service layer.

[0105] Embodiment 2

[0106] Please continue reading Figure 7 , showing a schematic diagram of the program module of the second embodiment of the financial statement recognition device of the present invention. In this embodiment, the financial statement recognition device 20 may include or be divided into one or more program modules, one or more program modules are stored in a storage medium and executed by one or more processors to complete the present invention and implement the above-mentioned financial statement recognition method. The program module referred to in the embodiment of the present invention refers to a series of computer program instruction segments that can perform specific functions, which is more suitable for describing the execution process of the financial statement recognition device 20 in the computer device 2 than the program itself. The following description will specifically introduce the functions of each program module of this embodiment:

[0107] The receiving module 200 is used to receive the financial report image to be processed.

[0108] Specifically, the financial report to be processed is a financial statement that needs to be identified, and the image of the financial report to be processed may be a scanned copy or a photo of the financial report to be processed. The photo of the financial report to be processed in the image of the financial report to be processed is relatively straight and neat.

[0109] The identification module 202 is used to identify the financial report image to be processed to obtain the financial report document contained in the financial report image to be processed, and the financial report document includes layout information, subject information and financial report data information.

[0110] Specifically, the financial report to be processed is identified through OCR technology, multiple field information is identified, and the multiple field information is analyzed to obtain the financial report type, financial report name, subject, and financial report data. The obtained financial report type can identify the layout of the financial report to be processed, and then match it with the built-in financial report template library. Different types of financial report layouts, or standard financial report templates are different, so the corresponding financial report type is obtained; the financial report contains type information, and after obtaining multiple field information through text recognition, it is matched with the fields of various standard financial report types to determine the financial report type; it can also be obtained through manual identification by the user. Subject information and financial report data information can be identified through key values, and a correlation is established. The corresponding financial report name is identified to obtain the main title, and the layout information includes the financial report layout, financial report type, and financial report name.

[0111] Exemplarily, the identification module 202 is further configured to:

[0112] The financial report to be processed is preprocessed to obtain a standard image.

[0113] Specifically, image input: For different image formats, there are different storage formats and different compression methods. Preprocessing: Mainly includes binarization, noise removal, tilt correction, etc. Binarization: Most of the pictures taken by the camera are color images. Color images contain a huge amount of information. The content of the picture can be simply divided into foreground and background. In order to make the computer recognize text faster and better, it is necessary to process the color image first so that the image only has foreground information and background information. The foreground information can be simply defined as black and the background information as white. This is the binarization image. Noise removal: For different documents, the definition of noise can be different. Denoising according to the characteristics of noise is called noise removal. Tilt correction: Since ordinary users are more casual when taking pictures of documents, the pictures taken will inevitably be tilted, which requires text recognition software to correct. After the above preprocessing, a standard picture is obtained.

[0114] Perform layout recognition on the standard image to obtain a corresponding financial report layout, wherein the layout information includes the financial report layout and the layout name.

[0115] Specifically, the process of dividing the document image into paragraphs and lines is called layout analysis. After the layout analysis, you can match it through the built-in templates. Different types of financial report templates are different, and get the financial report layout, such as general enterprises.

[0116] The standard image is subjected to character recognition to obtain a plurality of field information, wherein the field information includes the format name, subject information and financial report data information.

[0117] Specifically, character recognition can be performed through template matching and feature extraction. Due to the influence of factors such as text displacement, stroke thickness, broken strokes, adhesion, and rotation, the difficulty of feature extraction is greatly affected. Character recognition is performed on standard images through a text recognition model. Due to the limitations of shooting conditions, characters often stick together and break strokes, which greatly limits the performance of the recognition system. This requires the text recognition software to have a character cutting function. The extracted features can be recognized by a classifier, which classifies them and outputs which text the feature corresponds to. The spectral clustering algorithm, K nearest neighbor algorithm, and K value parameter space automatic search algorithm are used to determine the recognition of the segmented text and identify which text it is. After recognition, the layout name, subject information, and financial report data information are obtained, and the recognition results are corrected according to the relationship between the specific language context.

[0118] The layout of the financial report is restored according to the field information, and a verification is performed to obtain an identified financial report document.

[0119] Specifically, the recognized text is still arranged like the original document image, with the paragraphs, positions and order unchanged, and is output to a word document, pdf document, etc. This process is called layout recovery. After the recovery is completed, a financial report document is obtained. The fields of the account can be filled, and then the financial report data information can be directly filled into the target template.

[0120] Exemplarily, when the identified field content is not in the system financial report template library, the template can be automatically refined and supplemented.

[0121] The matching module 204 is used to input the format information and the subject information into a subject matching model and output a matching result, wherein the matching result includes the matching subjects of the to-be-processed financial report and the confidence level of the matching subjects.

[0122] Specifically, the layout information includes the financial report layout and the financial report name. The field information of the financial report layout, financial report name and subject information is input into the subject matching model as input data; after receiving the three parameters passed in, the subject matching model obtains the subject template set of the financial report template corresponding to the layout information, performs similarity calculations on the corresponding subjects, and generates subject matching results; the subject matching model outputs the matching results in the form of a dictionary, and the output content includes the matched subjects, confidence levels, and subjects to be selected. The financial report template is the financial report layout obtained after OCR recognition and analysis, such as general enterprises; the financial report name is the name of the financial report to be processed, such as cash flow statement, balance sheet, income statement, etc.; the subject is the subject after OCR technology recognition, such as notes receivable, etc. This step is to perform a rough recognition of financial report subjects, that is, to organize the text recognized by OCR.

[0123] Exemplarily, the matching module 204 is further configured to:

[0124] The format information and the subject information are input into a subject matching model, so as to obtain a subject template library matching the format information according to the format information through the subject matching model.

[0125] Specifically, the layout information includes the financial report layout and the financial report type. The field information of the financial report template, financial report name and account information is input into the account matching model as input data. After the account matching model receives the three passed in parameters, it obtains the account template set of the financial report template corresponding to the layout information. The account template set includes multiple standard accounts.

[0126] The subject information is segmented according to the granularity of characters to obtain character information corresponding to each subject, wherein the subject information includes a plurality of subjects.

[0127] Specifically, each subject in the subject information is segmented into at least two characters.

[0128] The character information corresponding to each of the subjects is matched according to a preset inverted index table to obtain a subject candidate set corresponding to each subject from the subject template library.

[0129] Specifically, the character information of each subject is matched with the subject template library according to the inverted index table. The inverted index table pre-stores the standard characters of the standard subjects obtained from the standard template, and each standard subject is numbered. As long as the standard characters corresponding to a certain standard template match the characters of the subject, the standard subject corresponding to the standard characters will be recalled. Similarity matching can be performed based on the character similarity algorithm of NLP technologies such as the longest common substring, edit distance algorithm, and cosine similarity algorithm. For example: when using the edit distance algorithm, calculate how many times the characters of the subject are edited to obtain the standard characters of the standard subject. The number of modifications is the similarity. The fewer the number, the more similar. The editing operation includes replacing one character with another, inserting a character, and deleting a character. For another example: when using the longest common substring, find the length of the same string in the characters of the subject and the standard characters of the standard subject in order. The longer the length, the greater the similarity. Then find the number corresponding to the standard subject in the inverted index table and associate it with the matching subject. The template corresponding to the number can be queried in the inverted index table in the order of the number to see if it contains character information of the subject. If it contains the subject, the standard subject corresponding to the number is recalled from the subject template library to obtain a subject candidate set.

[0130] Each of the subjects is matched with a standard subject in a subject candidate set corresponding to the subject, and a matching result is output, wherein the matching result includes a matching subject of the financial report to be processed and a confidence level of the matching subject.

[0131] Specifically, the similarity between each subject and the corresponding standard subject is calculated according to the similarity algorithm, and the confidence level is determined based on the similarity value. The matching result includes the matching subject, the confidence level of the matching subject, and the candidate subject. The matching subject is the financial report subject corresponding to the subject identified by the OCR technology; the confidence level: the accuracy of the matching, such as exact match, algorithm match, fuzzy match, manual intervention, etc.; the candidate subject: when an exact match and an algorithm match cannot be performed, the candidate matching subject is given for output and extracted during manual intervention.

[0132] Example of matching subjects: {

[0133] 'A': ['Jia', 1, []];

[0134] 'B': ['Yi', 2, []];

[0135] 'C': ['Bing', 3, ['big','small','many', 'few']];

[0136] 'D': ['Ding', 4, ['big','small','many', 'few']}

[0137] }

[0138] Explanation: A, B, C, and D are the financial report subjects identified by OCR. Jia, Yi, Bing, and Ding represent the matching subjects. 1, 2, 3, and 4 represent the identified confidence levels. When the identification is inaccurate (3, 4), Bing and Ding represent the candidate subjects with the highest similarity. The contents in ['big','small','many', 'few'] represent the candidate financial report subjects to be manually confirmed.

[0139] The determination and generation module 206 is configured to determine the target subject corresponding to the financial report to be processed from the matching subjects according to the confidence level, and generate a target financial report based on the target subject and the financial report data information.

[0140] Specifically, when the confidence level is high or medium-high, corresponding to 1 and 2 in the above example, the matching standard subject is used as the target subject; when the confidence level is low or medium-low, corresponding to 3 and 4 in the above example, the matching standard subject is used as the candidate subject, indicating that the subject does not fully match the standard subject. According to the confidence level between the subject and the standard subject, the recommended matching standard subjects are given, and then several fuzzy matching standard subjects (such as: ['big','small','many', 'few']) are obtained, and then verified manually to select the target subject. After selecting the target subject, the target subject is refilled into the financial report document to obtain the target financial report.

[0141] Embodiment III

[0142] See also Figure 8 , is a schematic diagram of the hardware architecture of the computer device of the third embodiment of the present invention. In this embodiment, the computer device 2 is a device that can automatically perform numerical calculations and / or information processing according to pre-set or stored instructions. The computer device 2 can be a rack server, a blade server, a tower server or a cabinet server (including an independent server, or a server cluster composed of multiple servers), etc. Figure 8 As shown, the computer device 2 at least includes, but is not limited to, a memory 21, a processor 22, a network interface 23, and a financial statement recognition device 20 that can be interconnected and communicated via a system bus.

[0143] In this embodiment, the memory 21 includes at least one type of computer-readable storage medium, and the readable storage medium includes flash memory, hard disk, multimedia card, card-type memory (for example, SD or DX memory, etc.), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, disk, optical disk, etc. In some embodiments, the memory 21 can be an internal storage unit of the computer device 2, such as a hard disk or memory of the computer device 2. In other embodiments, the memory 21 can also be an external storage device of the computer device 2, such as a plug-in hard disk equipped on the computer device 2, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), etc. Of course, the memory 21 can also include both the internal storage unit of the computer device 2 and its external storage device. In this embodiment, the memory 21 is generally used to store the operating system and various application software installed on the computer device 2, such as the program code of the financial statement recognition device 20 of the second embodiment. In addition, the memory 21 can also be used to temporarily store various types of data that have been output or are to be output.

[0144] The processor 22 may be a central processing unit (CPU), a controller, a microcontroller, a microprocessor, or other data processing chips in some embodiments. The processor 22 is generally used to control the overall operation of the computer device 2. In this embodiment, the processor 22 is used to run the program code stored in the memory 21 or process data, such as running the financial statement recognition device 20 to implement the financial statement recognition method of the first embodiment.

[0145] The network interface 23 may include a wireless network interface or a wired network interface, and the network interface 23 is generally used to establish a communication connection between the server 2 and other electronic devices. For example, the network interface 23 is used to connect the server 2 to an external terminal through a network, to establish a data transmission channel and a communication connection between the server 2 and the external terminal, etc. The network may be an intranet, the Internet, the Global System of Mobile communication (GSM), Wideband Code Division Multiple Access (WCDMA), 4G network, 5G network, Bluetooth, Wi-Fi and other wireless or wired networks. It should be pointed out that Figure 8 Computer device 2 is shown only with components 20 - 23 , but it should be understood that implementation of all of the components shown is not a requirement, and more or fewer components may alternatively be implemented.

[0146] In this embodiment, the financial statement recognition device 20 stored in the memory 21 can also be divided into one or more program modules, which are stored in the memory 21 and executed by one or more processors (processor 22 in this embodiment) to complete the financial statement recognition device of the present invention. The specific functions of the program modules 200-206 have been described in detail in the second embodiment and will not be repeated here. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covered on the display screen, or a button, trackball or touchpad set on the computer device housing, or an external keyboard, touchpad or mouse. Receive the user's operating instructions on the display screen of the computer device to achieve the steps of the financial statement recognition method, run the processor 22 according to the operating instructions, and execute the computer program stored on the computer readable storage medium 21. The steps of the financial statement recognition method here can be the steps in the financial statement recognition method of the above-mentioned embodiment one.

[0147] Embodiment 4

[0148] This embodiment also provides a computer-readable storage medium, which can be non-volatile or volatile. For example, flash memory, hard disk, multimedia card, card-type memory (for example, SD or DX memory, etc.), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, disk, optical disk, server, App application store, etc., which store computer programs, and the corresponding functions are realized when the programs are executed by the processor. The computer-readable storage medium of this embodiment stores a computer program, and when the computer program is executed by the processor, the processor executes the step device of the financial statement recognition method of the above-mentioned embodiment 1.

[0149] The serial numbers of the above embodiments of the present invention are only for description and do not represent the advantages or disadvantages of the embodiments.

[0150] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method.

[0151] The above are only preferred embodiments of the present invention, and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A method for identifying financial statements, characterized in that: include: Receive the pending financial report pictures of the pending financial report; Identify the financial report image to be processed to obtain a financial report document contained in the financial report image to be processed, wherein the financial report document includes layout information, subject information, and financial report data information; Inputting the format information and the subject information into a subject matching model, and outputting a matching result, wherein the matching result includes a matching subject of the to-be-processed financial report and a confidence level of the matching subject; Determine, from the matching subjects according to the confidence level, a target subject corresponding to the to-be-processed financial report, and generate a target financial report based on the target subject and the financial report data information; The step of identifying the financial report image to be processed to obtain the financial report document contained in the financial report image to be processed, wherein the financial report document includes layout information, subject information and financial report data information including: Preprocessing the financial report to be processed to obtain a standard image; Performing layout recognition on the standard image to obtain a corresponding financial report layout, wherein the layout information includes the financial report layout and the layout name; Performing character recognition on the standard image to obtain multiple field information, the field information including the format name, subject information and financial report data information; Recovering the layout of the financial report format according to the field information, and verifying it to obtain an identified financial report document; Wherein, the method further comprises: Get multiple financial report data; Performing data analysis on the target financial report and the financial report data to obtain analysis results; generating early warning information based on the analysis results; The step of inputting the format information and the subject information into a subject matching model and outputting a matching result, wherein the matching result includes the matching subject of the to-be-processed financial report and the confidence level of the matching subject, includes: Inputting the format information and the subject information into a subject matching model, so as to obtain a subject template library matching the format information according to the format information through the subject matching model; Segmenting the subject information according to the granularity of characters to obtain character information corresponding to each subject, wherein the subject information includes a plurality of subjects; Matching the character information corresponding to each of the subjects according to a preset inverted index table to obtain a subject candidate set corresponding to each subject from the subject template library; Each of the subjects is matched with a standard subject in a subject candidate set corresponding to the subject, and a matching result is output, wherein the matching result includes a matching subject of the financial report to be processed and a confidence level of the matching subject.

2. The method for identifying financial statements according to claim 1, characterized in that: Before matching the character information corresponding to each subject according to the preset inverted index to obtain the subject candidate set corresponding to each subject, the method includes: Get multiple standard subjects; A plurality of inverted index tables are established by using the first characters of the plurality of standard subjects as keywords, and the plurality of inverted index tables are stored in the subject template library.

3. The method for identifying financial statements according to claim 1, characterized in that: The matching of each of the subjects with the standard subjects in the subject candidate set corresponding to the subject, and outputting a matching result, wherein the matching result includes the matching subject of the to-be-processed financial report and the confidence level of the matching subject, includes: Calculate the similarity value between each subject and the standard subject in the subject candidate set corresponding to the subject according to the similarity algorithm; The standard subjects whose similarity values ​​are greater than a preset threshold are taken as matching subjects, and the confidence levels of the matching subjects are determined according to the similarity values ​​of the matching subjects.

4. The method for identifying financial statements according to claim 1, characterized in that: The method further comprises: The target financial report is stored in the blockchain.

5. A financial statement recognition device, characterized in that: include: A receiving module, used for receiving the financial report picture to be processed of the financial report to be processed; An identification module, used for identifying the financial report image to be processed to obtain a financial report document contained in the financial report image to be processed, wherein the financial report document includes format information, subject information and financial report data information; A matching module, used for inputting the format information and the subject information into a subject matching model, and outputting a matching result, wherein the matching result includes the matching subject of the to-be-processed financial report and the confidence level of the matching subject; A determination generation module, configured to determine a target subject corresponding to the to-be-processed financial report from the matching subjects according to the confidence level, and generate a target financial report based on the target subject and the financial report data information; The step of identifying the financial report image to be processed to obtain the financial report document contained in the financial report image to be processed, wherein the financial report document includes layout information, subject information and financial report data information including: Preprocessing the financial report to be processed to obtain a standard image; Performing layout recognition on the standard image to obtain a corresponding financial report layout, wherein the layout information includes the financial report layout and the layout name; Performing character recognition on the standard image to obtain multiple field information, the field information including the format name, subject information and financial report data information; Recovering the layout of the financial report format according to the field information, and verifying it to obtain an identified financial report document; Wherein, the device is also used for: Get multiple financial report data; Performing data analysis on the target financial report and the financial report data to obtain analysis results; generating early warning information based on the analysis results; The step of inputting the format information and the subject information into a subject matching model and outputting a matching result, wherein the matching result includes the matching subject of the to-be-processed financial report and the confidence level of the matching subject, includes: Inputting the format information and the subject information into a subject matching model, so as to obtain a subject template library matching the format information according to the format information through the subject matching model; Segmenting the subject information according to the granularity of characters to obtain character information corresponding to each subject, wherein the subject information includes a plurality of subjects; Matching the character information corresponding to each of the subjects according to a preset inverted index table to obtain a subject candidate set corresponding to each subject from the subject template library; Each of the subjects is matched with a standard subject in a subject candidate set corresponding to the subject, and a matching result is output, wherein the matching result includes a matching subject of the financial report to be processed and a confidence level of the matching subject.

6. A computer device, characterized in that: The computer device includes a memory and a processor. The memory stores a computer program that can be run on the processor. When the computer program is executed by the processor, the steps of the financial statement identification method according to any one of claims 1 to 4 are implemented.

7. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, which can be executed by at least one processor to enable the at least one processor to perform the steps of the financial statement identification method according to any one of claims 1 to 4.

Images