Intelligent financial report generation method and system

Abstract

The invention discloses an intelligent generation method and system for a financial report, and belongs to the technical field of data processing, and the method comprises the steps: setting a preset template based on a preset label, collecting financial records based on the preset template, generating to-be-imported data, constructing a knowledge summarization model, summarizing the to-be-imported data based on the knowledge summarization model, and generating summarized data; constructing a summary model, obtaining an abstract tag of the summarized data based on the summary model, matching the abstract tag with a preset tag to generate a matched tag, and generating matched data corresponding to the matched tag based on the summarized data; checking the matched data to generate standard texts, and combining the standard texts based on the matched labels to generate financial report texts; and performing risk analysis on the financial report text based on the matching tag, outputting a risk report, and completing report analysis of the financial record based on the financial report text and the risk report. According to the invention, the comprehensiveness and accuracy of generating the financial report can be improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of data processing, and in particular relates to an intelligent generation method and system for financial reports. Background Art

[0002] In the field of finance and accounting, the preparation and analysis of financial reports is one of the core tasks. Traditional financial report preparation mainly relies on manual operations, including multiple steps such as data collection, classification, calculation and interpretation. This method is not only time-consuming and labor-intensive, but also susceptible to human errors, such as data input errors and miscalculations. In addition, with the expansion of enterprise scale and the complexity of business, the amount of financial data has increased dramatically, and traditional manual processing methods have been unable to meet the needs of fast and accurate preparation of financial reports. In recent years, with the development of machine learning, natural language processing and data mining technology, intelligent financial report generation system has become a hot research topic.

[0003] Similar prior art includes a Chinese patent application with publication number CN112581247A, which discloses a financial report generation method and device, and relates to the field of computer technology. A specific implementation of the method includes receiving a financial report submission request, obtaining a corresponding financial report template according to the financial report type in the request; identifying a table mapping symbol in the financial report template, calling a preset data acquisition model, and generating data corresponding to the table mapping symbol; obtaining the position information of the table mapping symbol, clearing the table mapping symbol at the position information, writing the corresponding data, generating a financial report and outputting it. Thus, the implementation of the invention can solve the problem that the existing automatic and accurate financial reporting cannot be achieved. There is also a Chinese patent application with publication number CN109710684A, which discloses an intelligent financial report generation system based on a logical reasoning machine, including a financial information acquisition module, a financial database, a model building module, a warehousing database, a reasoning analysis module and a decision analysis module; the financial information acquisition module is connected to the financial database and the model building module respectively, the model building module is connected to the reasoning analysis module, and the reasoning analysis module is connected to the financial database, the warehousing database and the decision analysis module respectively. The invention provides an intelligent financial report generation system based on a logical reasoning engine, which classifies, organizes and analyzes financial data through a reasoning analysis module combined with a decision analysis module, statistically analyzes the decision content of the financial data, and automatically generates reports, making it easy for managers to understand the company's financial status based on the report, and intuitively displaying the changes in financial data, providing reliable decision-making content for the company's future financial development, and having the characteristics of high accuracy and high intelligence.

[0004] However, the above-mentioned prior art does not verify the data corresponding to the table mapping symbol, which reduces the integrity of the financial report and only performs decision analysis on numerical financial data. In actual situations, financial data is not only of numerical type, but also needs to be analyzed in combination with text-type financial data to improve the comprehensiveness of the financial report. Summary of the invention

[0005] To solve the above problems, the present invention provides a method and system for intelligently generating financial reports to solve the problems in the prior art.

[0006] In order to achieve the above-mentioned object of the invention, the present invention proposes an intelligent generation method of financial reports, comprising: S1: setting a preset template based on a preset tag, collecting financial records based on the preset template, generating data to be imported, building a knowledge aggregation model, and aggregating the data to be imported based on the knowledge aggregation model to generate aggregated data; S2: constructing a summary model, obtaining summary tags of the summary data based on the summary model, matching the summary tags with the preset tags to generate matching tags, and generating matching data corresponding to the matching tags based on the summary data; S3: Check the matching data to generate a standard text, and combine the standard text based on the matching tags to generate a financial report text; S4: Perform risk analysis on the financial report text based on the matching tags, output a risk report, and complete report analysis of the financial records based on the financial report text and the risk report.

[0007] Furthermore, the generating of the data to be imported comprises the following steps: Obtain the parsing structure of the preset template, obtain the attribute information corresponding to any character in the preset template based on the parsing structure, build a tag library based on the attribute information, determine whether there is response information between any element in the financial record and any attribute information in the tag library, and if so, extract conversion content from all the elements based on the response information, replace the characters corresponding to the attribute information with the conversion content, generate a new preset template, set the new preset template as the data to be imported, and repeat this step until all the elements in the financial record are extracted.

[0008] Furthermore, a parser is set, the preset template is input into the parser, all characters in the preset template are traversed, the parser extracts the data type of the character, encodes the data type, generates encoding information, combines all the encoding information and sets it as the parsing structure.

[0009] Furthermore, the determining whether there is response information between any element in the financial record and any attribute information in the tag library comprises the following steps: Based on the preset tag, the mapping relationship between the character and the attribute information is obtained, all the mapping relationships are combined to generate a mapping rule, all the financial records are converted into a preset format, and the data information contained in the preset format is set as the element. If the mapping rule exists between the element and any of the attribute information, it is determined that the response information exists.

[0010] Furthermore, generating summary data comprises the following steps: Building a financial learning model based on a neural network, inputting all the financial records into the financial learning model for training, and setting the trained financial learning model as the knowledge summary model; Inputting all the data to be imported into the knowledge aggregation model, extracting financial related data, performing data preprocessing on the financial related data, obtaining data features, classifying all the financial related data based on the data features, and generating multiple classification results; The financial-related data included in the classification results are combined based on financial relationship information to generate the summary data.

[0011] Furthermore, the step of inputting all the financial records into the financial learning model for training includes the following steps: Setting an analysis target, converting the financial record into a vector set based on special characters, dividing the vector set into a training set, a validation set, and a test set, training the financial learning model using the training set based on the analysis target, adjusting a performance parameter of the financial learning model based on the validation set, outputting an evaluation value based on the test set, and stopping training the financial learning model if the evaluation value is greater than a first preset value; The financial learning model outputs an analysis result corresponding to the analysis target, and sets the analysis result as the financial relationship information.

[0012] Furthermore, generating the matching data corresponding to the preset tag based on the summary data includes the following steps: The summary model generates a frequency histogram corresponding to all phrases in the summary data, obtains the part-of-speech type corresponding to the phrase, and combines all the phrases based on the frequency histogram and the part-of-speech type to generate a plurality of first combination words; Setting a financial word library, extracting semantic phrases of the first combination of words from the financial word library and setting them as second combination of words, calculating the association probability between any second combination of words and the first combination of words, accumulating the association probabilities to generate a first value, and if the first value is greater than or equal to a second preset value, setting the second combination of words with the largest association probability as the summary label, otherwise, retaining the first combination of words and setting it as the summary label; Obtaining a semantic vector angle between the summary tag and the preset tag, and if the semantic vector angle is less than or equal to a third preset value, determining that the summary tag matches the preset tag, and setting the summary tag as the matching tag; The summary data is split into multiple sub-texts, the association degree between the sub-text and any matching tag is obtained, all the sub-texts are clustered based on the association degree and the matching tag to generate multiple clustering clusters, and the sub-text included in any clustering cluster is set as the matching data corresponding to the matching tag.

[0013] Furthermore, the generating of the financial report text comprises the following steps: Check whether any of the matching tags has the matching data, if so, use a text generator to convert the matching data into the standard text, if not, set missing information; The standard text, the missing information and the matching tags are matched one by one and combined to generate the financial report text.

[0014] Furthermore, completing the report analysis of the financial records based on the financial report text and the risk report includes the following steps: Extracting the numerical information contained in the financial report text, and obtaining connecting phrases of the numerical information, calculating the mathematical relationship between the numerical information based on the connecting phrases, combining the mathematical relationship, the connecting phrases and the numerical information based on natural language processing technology to generate a summary text, setting the summary text as the risk report, and writing the risk report into the financial report text; If the mathematical relationship does not meet the preset financial standard, a warning message is generated, and the warning message and the financial report text are sent to the management personnel to complete the report analysis of the financial record.

[0015] The present invention also provides a system for intelligently generating financial reports, which is used to implement the above-mentioned method for intelligently generating financial reports. The system mainly includes: An integration module, setting a preset template based on a preset tag, collecting financial records based on the preset template, generating data to be imported, building a knowledge aggregation model, aggregating the data to be imported based on the knowledge aggregation model, and generating aggregated data; A matching module, constructing a summary model, obtaining a summary tag of the summary data based on the summary model, matching the summary tag with the preset tag to generate a matching tag, and generating matching data corresponding to the matching tag based on the summary data; A verification module verifies the matching data, generates a standard text, and combines the standard text based on the matching tags to generate a financial report text; The analysis module performs risk analysis on the financial report text based on the matching tags, outputs a risk report, and completes report analysis of the financial record based on the financial report text and the risk report.

[0016] Compared with the prior art, the beneficial effects of the present invention are at least as follows: The present invention first collects financial records through preset templates, which can quickly and accurately generate data to be imported and reduce the processing process of unnecessary financial records. Then, the data to be imported is classified and summarized through a knowledge summary model to generate summary data, which can improve the usability of the data to be imported. Finally, the summary model is used to match the summary data with matching tags to generate matching data, which can improve the precision of the financial report text.

[0017] The present invention also performs risk analysis on the financial report text by calculating mathematical relationships and outputs warning information, thereby improving the overall quality of the financial report text and realizing intelligent analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 A flowchart of the steps of an intelligent generation method of financial reports according to the present invention; Figure 2 This is a structural diagram of an intelligent generation system of financial reports according to the present invention. DETAILED DESCRIPTION

[0019] In order to make the purpose, technical solution 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.

[0020] It is understood that the terms "first", "second", etc. used in this application may be used herein to describe various elements, but unless otherwise specified, these elements are not limited by these terms. These terms are only used to distinguish a first element from another element. For example, without departing from the scope of this application, a first xx script may be referred to as a second xx script, and similarly, a second xx script may be referred to as a first xx script.

[0021] like Figure 1 As shown, a method for intelligently generating financial reports includes: S1: Set a preset template based on a preset tag, collect financial records based on the preset template, generate data to be imported, build a knowledge aggregation model, aggregate the data to be imported based on the knowledge aggregation model, and generate aggregated data.

[0022] Specifically, in this embodiment, the preset template refers to the system template data used to generate financial reports in the financial management system. The data types in the preset template include but are not limited to numeric types, character types, and Boolean types, etc., wherein the preset label refers to the label information contained in the preset template, such as company profile, important accounting policies, financial expenditures, etc. Financial records refer to financial information written and stored in the financial management system, including various data types, and the data to be imported refers to the data information in the financial records that is needed to generate financial reports. The knowledge summary model refers to a neural network model for data classification, and the summary data refers to the data information generated after pre-analysis of the data to be imported, wherein there is financial correlation between the data information contained in the summary data.

[0023] S2: Build a summary model, obtain summary labels of the summary data based on the summary model, match the summary labels with preset labels, generate matching labels, and generate matching data corresponding to the matching labels based on the summary data.

[0024] Specifically, in this embodiment, the summary model is a model for text summarization, which includes mathematical statistics, natural language processing and other technologies. The summary tag refers to the tag information generated after summarizing the summary data. If the summary tag matches the preset tag, it means that the corresponding summary data contains matching data associated with the preset template. Therefore, the corresponding matching data is combined by matching tags.

[0025] S3: Check the matching data, generate standard text, combine the standard text based on the matching tags, and generate the financial report text.

[0026] Specifically, in this embodiment, in order to perform quality control on the financial report text and reduce subsequent corrections, it is necessary to verify the matching data, and generate standard text through the verified matching data. The standard text is classified and combined through matching tags to generate the financial report text, wherein the financial report text refers to the text information corresponding to the financial report, including data types such as numeric type and character type.

[0027] S4: Perform risk analysis on the financial report text based on matching tags, output a risk report, and complete report analysis of financial records based on the financial report text and the risk report.

[0028] Specifically, in this embodiment, in order to timely remind risk points of the financial report text, it is necessary to perform risk analysis on the financial report text and output a corresponding risk report, and complete the report analysis of all the above financial records in combination with the risk report and the financial report text.

[0029] As a preferred technical solution of the present invention, generating the data to be imported includes the following steps: Get the parsing structure of the preset template, get the attribute information corresponding to any character in the preset template based on the parsing structure, build a tag library based on the attribute information, determine whether there is response information between any element in the financial record and any attribute information in the tag library, if so, extract the conversion content in all elements based on the response information, and replace the characters corresponding to the attribute information with the conversion content, generate a new preset template, set the new preset template as the data to be imported, and repeat this step until all elements in the financial record are extracted.

[0030] Specifically, in this embodiment, the parsing structure refers to the coding structure generated by the parser, which can be used to read and understand the structure of the preset template. The financial record can be converted into the XBRL markup language through the parsing structure corresponding to the preset template, wherein the financial record refers to the record information stored in the financial management system. XBRL (eXtensible Business Reporting Language) is a markup language based on XML (eXtensible Markup Language) and is specially designed for electronic exchange of business report data. Attribute information refers to the importance of the context information used to describe the characters in the parsing structure, which can match and extract the key information of the corresponding data. Response information refers to the response process. If there is response information, it means that the element in the financial record belongs to the data point corresponding to the attribute information, and the element is set as the conversion content, wherein the conversion content refers to the data information corresponding to the parsing structure in the financial record. Therefore, the character corresponding to the attribute information in the preset template is replaced with the conversion content, and then the data information contained in the replaced new preset template is set as the data to be imported. Through this step, the necessary data information in the financial record can be screened out, reducing the complexity of subsequent data analysis.

[0031] As a preferred technical solution of the present invention, a parser is set, a preset template is input into the parser, all characters in the preset template are traversed, the parser extracts the data type of the character, encodes the data type, generates encoding information, combines all the encoding information and sets it as a parsing structure.

[0032] Specifically, in this embodiment, the parser refers to a software tool that can read and interpret data files in a specific format and convert them into structured data that can be processed by the program. The parser encodes the characters corresponding to the preset template to generate encoding information, and generates a parsing structure by combining the encoding information.

[0033] As a preferred technical solution of the present invention, judging whether there is response information between any element in the financial record and any attribute information in the tag library includes the following steps: Based on the preset tags, the mapping relationship between characters and attribute information is obtained, all mapping relationships are combined to generate mapping rules, all financial records are converted into a preset format, and the data information contained in the preset format is set as an element. If there is a mapping rule between the element and any attribute information, it is determined that there is response information.

[0034] Specifically, in this embodiment, the mapping relationship refers to the corresponding relationship between the attribute information in the preset template and the specific data point corresponding to the character in the financial record. The mapping rule refers to the set of mapping relationships, which is used to define the rules for the relationship between the attribute information and the character. The preset format refers to the format corresponding to the XBRL markup language converted into the financial record, where the element refers to the format data contained in the preset format. If there is a mapping rule between the element and any attribute information, it means that the element belongs to the data point corresponding to the attribute information, that is, the element is data information that needs to be repeatedly extracted, and therefore, it is determined that there is response information.

[0035] As a preferred technical solution of the present invention, generating summary data includes the following steps: A financial learning model is constructed based on a neural network, all financial records are input into the financial learning model for training, and the trained financial learning model is set as a knowledge summary model.

[0036] Input all the data to be imported into the knowledge aggregation model, extract the financial related data, perform data preprocessing on the financial related data, obtain the data features, classify all the financial related data based on the data features, and generate multiple classification results.

[0037] The financial-related data included in the classification results are combined based on the financial relationship information to generate summary data.

[0038] Specifically, in this embodiment, the financial learning model composed of a neural network can perform learning and training on financial records to generate a knowledge summary model for classification.

[0039] The data to be imported contains financial-related data, i.e., financial-related data. Data preprocessing refers to cleaning and formatting the input financial-related data to ensure the quality and consistency of the data. Data features refer to the feature information generated after extracting elements from financial-related data, and the relative contribution between elements is calculated through data features, and the financial-related data is classified by the relative contribution. For example, the Markov logic network is first applied to define the logic formula, and then for each element (data feature), its contribution to other factors, i.e., the relative contribution, is calculated. Finally, the data features with a relative contribution greater than the first threshold are classified into one category to generate multiple classification results.

[0040] Financial relationship information refers to the mutual relationships corresponding to the relationship elements in the financial related data using natural language processing, such as causal relationships, correlations, hierarchical relationships, and logical relationships. Based on the results of the financial relationship information, the various financial related data contained in the classification results are sequentially combined to generate summary data.

[0041] As a preferred technical solution of the present invention, inputting all financial records into the financial learning model for training includes the following steps: Set the analysis goal, convert the financial records into a vector set based on special characters, divide the vector set into a training set, a validation set and a test set, use the training set to train the financial learning model based on the analysis goal, adjust the performance parameters of the financial learning model based on the validation set, output an evaluation value based on the test set, and stop training the financial learning model if the evaluation value is greater than a first preset value.

[0042] The financial learning model outputs analysis results corresponding to the analysis targets, and sets the analysis results as financial relationship information.

[0043] Specifically, in this embodiment, the vector set refers to a set corresponding to a semantic vector, wherein the semantic vector refers to a vector generated by semantic analysis of data information after the financial record is segmented by special characters, and the special characters can be set to characters corresponding to punctuation marks. The learning and training process of the financial learning model can be optimized through the training set, the validation set, and the test set. The evaluation value refers to the final performance value of the financial learning model, including but not limited to the accuracy, recall rate, etc.

[0044] The analysis result refers to the relationship result generated by the financial learning model through analyzing the mutual relationship through natural language processing technology. Therefore, the analysis result is set as financial relationship information.

[0045] As a preferred technical solution of the present invention, generating matching data corresponding to preset tags based on summary data includes the following steps: The summary model generates a frequency histogram corresponding to all phrases in the summary data, obtains the part-of-speech type corresponding to the phrase, and combines all phrases based on the frequency histogram and the part-of-speech type to generate a plurality of first combination words.

[0046] A financial vocabulary is set, a semantic phrase of the first combination of words is extracted from the financial vocabulary and set as a second combination of words, an association probability between any second combination of words and the first combination of words is calculated, the association probabilities are accumulated to generate a first numerical value, and if the first numerical value is greater than or equal to a second preset value, the second combination of words with the largest association probability is set as a summary label, otherwise, the first combination of words is retained and set as the summary label.

[0047] The semantic vector angle between the summary tag and the preset tag is obtained. If the semantic vector angle is less than or equal to a third preset value, it is determined that the summary tag matches the preset tag, and the summary tag is set as a matching tag.

[0048] The summary data is split into multiple sub-texts, the correlation degree between the sub-text and any matching label is obtained, all the sub-texts are clustered based on the correlation degree and the matching label to generate multiple clusters, and the sub-text contained in any cluster is set as the matching data corresponding to the matching label.

[0049] Specifically, in this embodiment, the summary model first extracts each phrase in the summary data, and draws a histogram based on the frequency of occurrence of the same phrase, i.e., a frequency histogram. The part-of-speech type refers to the information corresponding to the part-of-speech type, including but not limited to nouns, verbs, and adjectives. The first combination word is a compound word list, firstly, all phrases are marked according to the word type, and phrases with higher frequencies are selected as candidate phrases according to the frequency histogram, and then the candidate phrases with relevant part-of-speech types are combined based on language rules to generate compound words, and the rationality of the compound words is evaluated through semantic analysis, and multiple compound words that meet the rationality are set as a compound word list, i.e., the first combination word.

[0050] The financial word library refers to a collection of finance-related phrases, the semantic word library refers to a word in the financial word library that has semantic relevance to the first word combination, and the association probability refers to the result of normalizing the TF-IDF value of the second word combination in the first word combination, where TF-IDF (Term Frequency-Inverse Document Frequency) is a commonly used weighting technology for information retrieval and text mining, reflecting the importance of a second word combination to the first word combination. Since there may be multiple second words combination, the corresponding association probability is also multiple.

[0051] If the first numerical value is greater than or equal to the second preset value, it means that the semantic relevance between the first combination of words and the financial vocabulary is large, and it contains rich phrase information. Therefore, the first combination of words is replaced with the second combination of words with the highest association probability and used as the summary label. If the first numerical value is less than the second preset value, it means that the semantic relevance between the first combination of words and the financial vocabulary is low, and it may belong to the summary text of information summary. Therefore, the first combination of words is directly set as the summary label. For example, the first combination of words is "Apple Sales Overview", where "apple" may have multiple meanings, which can be the name of a fruit or the name of a company, and the semantic phrases corresponding to the first combination of words in the financial vocabulary are "Apple Company", "Apple Company Turnover", etc., then the second combination of words with the highest association probability "Apple Company Turnover" is selected as the summary label. Through the above steps, the diversity of summary labels can be improved and the ambiguity of summary labels can be reduced.

[0052] Extract multiple phrases to form information corresponding to the summary tag, wherein the multiple phrases can be phrases with higher frequencies in the frequency histogram and different parts of speech types.

[0053] The semantic vector angle refers to the vector angle between the semantic vector of the summary tag and the semantic vector of the preset tag. The smaller the semantic vector angle, the more similar the two semantic vectors are, and the summary tag and the preset tag match each other. Therefore, the summary tag that matches the preset tag is extracted and set as the matching tag.

[0054] The degree of association refers to the numerical value of the correlation, which can be calculated through semantic analysis. Subtexts with the same matching label and a high degree of association are clustered into a cluster. Therefore, multiple clusters are generated, and the subtexts contained in the clusters are combined and set as matching data corresponding to the matching label.

[0055] As a preferred technical solution of the present invention, generating a financial report text includes the following steps: Check if there is matching data for any matching tag. If so, use the text generator to convert the matching data into standard text. If not, set missing information.

[0056] The standard text, missing information and matching labels are matched one by one and combined to generate the financial report text.

[0057] Specifically, in this embodiment, the text generator is an algorithm that can convert matching data into accurate standard text based on different technologies and methods, including rule-based systems, machine learning, and deep learning models. Missing information refers to the reminder information set by the lack of matching data in the matching tag.

[0058] Based on the format of the preset template, the matching tag, the standard text corresponding to the matching tag, and the missing information are recombined and written into the format corresponding to the preset template to generate the financial report text. One-to-one correspondence means that the standard text corresponds to the matching tag, and at the same time, the missing information corresponds to the matching tag.

[0059] As a preferred technical solution of the present invention, the report analysis of financial records based on the financial report text and the risk report includes the following steps: Extract the numerical information contained in the financial report text, obtain the connecting phrases of the numerical information, calculate the mathematical relationship between the numerical information based on the connecting phrases, combine the mathematical relationship, connecting phrases and numerical information based on natural language processing technology to generate a summary text, set the summary text as a risk report, and write the risk report into the financial report text.

[0060] If the mathematical relationship does not meet the preset financial standards, a warning message is generated, and the warning message and financial report text are sent to management personnel to complete the report analysis of financial records.

[0061] Specifically, in this embodiment, numerical information refers to numerical data information, and connecting phrases refer to phrases that have an explanatory relationship with numerical information. For example, if "total assets are 100" in the financial report text, the numerical information is "100" and the connecting phrase is "total assets". Mathematical relationship refers to the changing relationship between numerical information generated by mathematical analysis formulas, including but not limited to addition relationship, subtraction relationship, percentage relationship and growth rate. For example, if "the turnover of No. 1 is 100 and the turnover of No. 2 is 50" in the financial report text, there is a mathematical relationship between "100" and "50" in the numerical information. Summary text refers to data analysis text generated based on mathematical relationship, connecting phrases and numerical information. For example, if the mathematical relationship is a subtraction relationship, the summary text is "the turnover decreases by (|50-100|) / 100=50%".

[0062] Financial standards refer to numerical standards related to finance, and warning information refers to the risk points used to promptly remind financial report texts. For example, the warning information can be set as "turnover does not meet the standard." Warning information and financial report texts containing risk reports are sent to management personnel to improve the integrity of report analysis.

[0063] like Figure 2 As shown, the present invention also provides an intelligent generation system for financial reports, which is used to implement the above-mentioned intelligent generation method for financial reports. The system mainly includes: The integration module sets a preset template based on a preset tag, collects financial records based on the preset template, generates data to be imported, builds a knowledge aggregation model, aggregates the data to be imported based on the knowledge aggregation model, and generates aggregated data.

[0064] The matching module builds a summary model, obtains summary tags of the summary data based on the summary model, matches the summary tags with preset tags, generates matching tags, and generates matching data corresponding to the matching tags based on the summary data.

[0065] The verification module verifies the matching data, generates standard text, combines the standard text based on matching tags, and generates financial report text.

[0066] The analysis module performs risk analysis on the financial report text based on matching tags, outputs a risk report, and completes report analysis of financial records based on the financial report text and the risk report.

[0067] It should be understood that, although each step in the flow chart of each embodiment of the present invention is shown in sequence according to the indication of the arrow, these steps are not necessarily performed in sequence according to the order indicated by the arrow. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be performed in other orders. Moreover, at least a portion of the steps in each embodiment may include a plurality of sub-steps or a plurality of stages, and these sub-steps or stages are not necessarily performed at the same time, but can be performed at different times, and the execution order of these sub-steps or stages is not necessarily performed in sequence, but can be performed in turn or alternately with at least a portion of other steps or sub-steps or stages of other steps.

[0068] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the above-mentioned program can be stored in a non-volatile computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0069] The technical features of the above-mentioned embodiments can be combined arbitrarily. In order to make the description concise, not all possible combinations of the technical features in the above-mentioned embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0070] The above embodiments only express several implementation methods of the present invention, and the description is relatively specific and detailed, but it cannot be understood as limiting the scope of the present invention. It should be pointed out that for ordinary technicians in this field, several modifications and improvements can be made without departing from the concept of the present invention, which all belong to the protection scope of the present invention.

[0071] The above are only preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for intelligently generating financial reports, characterized in that: The method comprises the following steps: S1: setting a preset template based on a preset tag, collecting financial records based on the preset template, generating data to be imported, building a knowledge aggregation model, and aggregating the data to be imported based on the knowledge aggregation model to generate aggregated data; S2: constructing a summary model, obtaining summary tags of the summary data based on the summary model, matching the summary tags with the preset tags to generate matching tags, and generating matching data corresponding to the matching tags based on the summary data; S3: Check the matching data to generate a standard text, and combine the standard text based on the matching tags to generate a financial report text; S4: Perform risk analysis on the financial report text based on the matching tags, output a risk report, and complete report analysis of the financial records based on the financial report text and the risk report.

2. The method according to claim 1, characterized in that The generating of the data to be imported comprises the following steps: Obtain the parsing structure of the preset template, obtain the attribute information corresponding to any character in the preset template based on the parsing structure, build a tag library based on the attribute information, determine whether there is response information between any element in the financial record and any attribute information in the tag library, and if so, extract conversion content from all the elements based on the response information, replace the characters corresponding to the attribute information with the conversion content, generate a new preset template, set the new preset template as the data to be imported, and repeat this step until all the elements in the financial record are extracted.

3. The method according to claim 2, characterized in that A parser is set, the preset template is input into the parser, all characters in the preset template are traversed, the parser extracts the data type of the character, encodes the data type, generates encoding information, combines all the encoding information and sets it as the parsing structure.

4. The method according to claim 2, characterized in that: The determining whether there is response information between any element in the financial record and any attribute information in the tag library comprises the following steps: Based on the preset tag, the mapping relationship between the character and the attribute information is obtained, all the mapping relationships are combined to generate a mapping rule, all the financial records are converted into a preset format, and the data information contained in the preset format is set as the element. If the mapping rule exists between the element and any of the attribute information, it is determined that the response information exists.

5. The method according to claim 1, characterized in that: Generating summary data comprises the following steps: Building a financial learning model based on a neural network, inputting all the financial records into the financial learning model for training, and setting the trained financial learning model as the knowledge summary model; Inputting all the data to be imported into the knowledge aggregation model, extracting financial related data, performing data preprocessing on the financial related data, obtaining data features, classifying all the financial related data based on the data features, and generating multiple classification results; The financial-related data included in the classification results are combined based on financial relationship information to generate the summary data.

6. The method according to claim 5, characterized in that The step of inputting all the financial records into the financial learning model for training comprises the following steps: Setting an analysis target, converting the financial record into a vector set based on special characters, dividing the vector set into a training set, a validation set, and a test set, training the financial learning model using the training set based on the analysis target, adjusting a performance parameter of the financial learning model based on the validation set, outputting an evaluation value based on the test set, and stopping training the financial learning model if the evaluation value is greater than a first preset value; The financial learning model outputs an analysis result corresponding to the analysis target, and sets the analysis result as the financial relationship information.

7. The method according to claim 1, characterized in that Generating the matching data corresponding to the preset tag based on the summary data comprises the following steps: The summary model generates a frequency histogram corresponding to all phrases in the summary data, obtains the part-of-speech type corresponding to the phrase, and combines all the phrases based on the frequency histogram and the part-of-speech type to generate a plurality of first combination words; Setting a financial word library, extracting semantic phrases of the first combination of words from the financial word library and setting them as second combination of words, calculating the association probability between any second combination of words and the first combination of words, accumulating the association probabilities to generate a first value, and if the first value is greater than or equal to a second preset value, setting the second combination of words with the largest association probability as the summary label, otherwise, retaining the first combination of words and setting it as the summary label; Obtaining a semantic vector angle between the summary tag and the preset tag, and if the semantic vector angle is less than or equal to a third preset value, determining that the summary tag matches the preset tag, and setting the summary tag as the matching tag; The summary data is split into multiple sub-texts, the association degree between the sub-text and any matching tag is obtained, all the sub-texts are clustered based on the association degree and the matching tag to generate multiple clustering clusters, and the sub-text included in any clustering cluster is set as the matching data corresponding to the matching tag.

8. The method according to claim 7, characterized in that Generating the financial report text comprises the following steps: Check whether any of the matching tags has the matching data, if so, use a text generator to convert the matching data into the standard text, if not, set missing information; The standard text, the missing information and the matching tags are matched one by one and combined to generate the financial report text.

9. The method according to claim 8, characterized in that The report analysis of the financial records based on the financial report text and the risk report comprises the following steps: Extracting the numerical information contained in the financial report text, and obtaining connecting phrases of the numerical information, calculating the mathematical relationship between the numerical information based on the connecting phrases, combining the mathematical relationship, the connecting phrases and the numerical information based on natural language processing technology to generate a summary text, setting the summary text as the risk report, and writing the risk report into the financial report text; If the mathematical relationship does not meet the preset financial standard, a warning message is generated, and the warning message and the financial report text are sent to the management personnel to complete the report analysis of the financial record.

10. An intelligent generation system of financial reports, used to implement the method according to any one of claims 1 to 9, characterized in that: The system includes the following modules: An integration module, setting a preset template based on a preset tag, collecting financial records based on the preset template, generating data to be imported, building a knowledge aggregation model, aggregating the data to be imported based on the knowledge aggregation model, and generating aggregated data; A matching module, constructing a summary model, obtaining a summary tag of the summary data based on the summary model, matching the summary tag with the preset tag to generate a matching tag, and generating matching data corresponding to the matching tag based on the summary data; A verification module verifies the matching data, generates a standard text, and combines the standard text based on the matching tags to generate a financial report text; The analysis module performs risk analysis on the financial report text based on the matching tags, outputs a risk report, and completes report analysis of the financial record based on the financial report text and the risk report.

Images