A knowledge graph-based on-site audit assistance method and device, equipment and medium

Abstract

The application discloses a knowledge graph-based on-site audit auxiliary method and device, equipment and medium, and relates to the technical field of automatic auditing. The application models the audit system, automatically obtains the corresponding audit model based on the audit project, and does not need the audit personnel to query, and then automatically obtains the suspicious data based on the audit model, so that the audit personnel only needs to check the suspicious data on site, reduces the workload of on-site checking, further generates the audit record based on the knowledge graph, and generates the audit draft based on the knowledge and experience, without the need for the audit personnel to write. The application improves the automation degree of the audit work, and further improves the audit efficiency.

Description

Technical Field

[0001] This application relates to the field of automated auditing technology, and in particular to a knowledge graph-based on-site auditing assistance method, apparatus, equipment, and medium. Background Technology

[0002] With the development of auditing technology, digital tools are continuously improving the efficiency of auditing work, especially in off-site audits, where digital methods can effectively shorten on-site operation time. However, existing auditing systems still have many shortcomings in on-site evidence collection and audit record generation. For example, auditors need to check regulatory basis, prepare audit records and reports, copy or scan key information, photograph on-site inventory evidence, and record audio and video interviews during the review process. These operations lack the support of digital and information technology, resulting in low efficiency and insufficient accuracy.

[0003] Especially in the internal auditing process of commercial banks, auditors need to quickly and accurately locate the required policy documents within a vast database of regulations. Existing manual auditing methods rely on the auditors' specialized skills; manually searching for regulations is time-consuming, labor-intensive, and highly subjective, easily leading to misunderstandings. Furthermore, the large volume and frequent updates of policy documents demand high skill levels from auditors, requiring them to be proficient in all relevant policy content. These issues result in low audit efficiency, failing to meet the demands of modern auditing work.

[0004] Furthermore, traditional digital auditing methods cannot achieve full coverage of the project process and cannot effectively audit incomplete data. When business data is not fully integrated, some audit procedures cannot be executed, ultimately affecting the quality of the project audit. Many audit tasks still need to be carried out on-site, resulting in lower audit efficiency and costs, and failing to meet the audit supervision system requirements of the new audit standards. Summary of the Invention

[0005] The purpose of this application is to provide a knowledge graph-based on-site audit assistance method, device, equipment, and medium to improve the automation level of audit work and thus improve audit efficiency.

[0006] To achieve the above objectives, this application provides the following solution.

[0007] Firstly, this application provides a knowledge graph-based on-site audit assistance method, including:

[0008] Determine the audit model corresponding to the audit project; the audit model is created based on the audit system corresponding to the audit project.

[0009] Based on the audit model, the audit data corresponding to the audit project is reviewed to identify suspicious data;

[0010] On-site verification was used to determine the evidence photos, verification results, and evidence collection records for each suspicious data point.

[0011] An evidence collection form is established based on the verified suspicious data, the evidence photos and records collected.

[0012] A knowledge graph is used to analyze the text data in the evidence collection table and generate audit records; the knowledge graph stores different audit standards and different historical audit records; the text data includes suspicious data and its evidence collection photos and records.

[0013] Based on the audit records and experiential knowledge, audit working papers are automatically generated; the experiential knowledge includes an audit question bank and a case bank.

[0014] Optionally, a knowledge graph can be used to analyze the textual data in the evidence collection form and generate audit records, specifically including:

[0015] Calculate the similarity between text data i and each audit standard and each historical audit record in the knowledge graph to determine the behavioral classification result of text data i, wherein the behavioral classification result is compliant, non-compliant or suspected problem; i = 1, 2, ..., I, where I is the number of text data in the evidence table;

[0016] Using an intent recognition model, the intent type of text data i is identified; the intent type is a request, an inquiry, or a statement.

[0017] Structured information is extracted from text data i using a natural language recognition model; the structured information includes: time, location, people, and events;

[0018] Based on the behavior classification results, intent type, and structured information of text data i, an audit record for text data i is generated using a preset template.

[0019] Optionally, the similarity between text data i and each audit standard and each historical audit record in the knowledge graph is calculated to determine the behavioral classification result of text data i, specifically including:

[0020] The audit standard with the highest similarity to text data i in the knowledge graph is identified as the target audit standard.

[0021] Determine whether text data i meets the target audit criteria, and obtain the first audit result for text data i;

[0022] Identify the historical audit record in the knowledge graph that has the highest similarity to text data i, and use it as the target audit record;

[0023] The audit results in the target audit record are determined as the second audit result of text data i;

[0024] When both the first audit result and the second audit result are compliant, the behavior classification result is determined to be compliant;

[0025] When both the first audit result and the second audit result are non-compliant, the behavior classification result is determined to be non-compliant;

[0026] When one of the first audit result and the second audit result is compliant and the other is non-compliant, the behavior classification result is determined to be a suspected problem.

[0027] Optionally, the knowledge graph is established in the following manner;

[0028] Obtain a database set; the sources of the knowledge text data in the database set include: audit notices, question databases, policy databases, and case databases;

[0029] Preprocess the knowledge text data in the database set;

[0030] Entity recognition and relation extraction are performed on each preprocessed knowledge text data to obtain an entity relation set;

[0031] Merge the relationships of the same entity from different sources in the entity relationship set to obtain the entity relationship set after relationship merging.

[0032] A knowledge graph is constructed based on the set of entity relationships.

[0033] Optionally, preprocessing includes: deduplication, data cleaning, and standardization;

[0034] The deduplication method is as follows: determine that two pieces of knowledge text data whose absolute value of the difference between hash function values ​​in the database set is less than a hash threshold are duplicate data, and delete one of them;

[0035] The data cleaning method is as follows: remove invalid characters from each knowledge text data and correct spelling errors in each knowledge text data.

[0036] Optionally, relationships from different sources within the entity relationship set can be merged to obtain a merged entity relationship set, specifically including:

[0037] Calculate the cosine similarity between two entities from different sources within a set of entity relations;

[0038] Determine whether two entities with a cosine similarity greater than a similarity threshold have the same relationship to obtain the first judgment result;

[0039] If the first judgment result is yes, then merge the two entities and their relationship with a cosine similarity greater than the similarity threshold;

[0040] If the first judgment result is negative, then determine whether the relationship between two entities with a cosine similarity greater than the first similarity threshold is conflicting, and obtain the second judgment result;

[0041] If the second judgment result is negative, then retain the two entities and their relationship with a cosine similarity greater than the first similarity threshold.

[0042] If the second judgment result is yes, delete the two entities and their relationship whose cosine similarity is greater than the first similarity threshold.

[0043] Optionally, based on the audit records and experiential knowledge, audit working papers can be automatically generated, specifically including:

[0044] Using natural language processing technology, audit question records in the audit question database with a similarity greater than a second similarity threshold are identified as first reference audit question records, and case records in the case database with a similarity greater than a third similarity threshold are identified as first reference case records.

[0045] Using data mining techniques, audit issue records in the audit issue database that have potential connections with the audit issue records are identified as second reference audit issue records, and case records in the case database that have potential connections with the audit issue records are identified as second reference case records.

[0046] Key information is extracted from each first reference audit issue record, each second reference audit issue record, each first case record, and each second case record; the key information includes: problem discovery, solutions, and risk points.

[0047] Secondly, this application provides a knowledge graph-based on-site audit assistance device, which applies the aforementioned knowledge graph-based on-site audit assistance method. The knowledge graph-based on-site audit assistance device includes:

[0048] The audit model determination module is used to determine the audit model corresponding to the audit project; the audit model is created based on the audit system corresponding to the audit project.

[0049] The suspicious data identification module is used to review the audit data corresponding to the audit project based on the audit model and identify suspicious data.

[0050] The evidence collection module is used to determine the evidence photos, verification results, and evidence collection records for each suspicious data point through on-site verification.

[0051] The evidence collection form creation module is used to create an evidence collection form based on the suspicious data that has been verified as true, as well as the evidence collection photos and records.

[0052] The audit record generation module is used to analyze the text data in the evidence collection table using a knowledge graph to generate audit records; the knowledge graph stores different audit standards and different historical audit records; the text data includes suspicious data and its evidence collection photos and records.

[0053] The audit working paper generation module is used to automatically generate audit working papers based on the audit records and experiential knowledge; the experiential knowledge includes an audit question bank and a case bank.

[0054] Thirdly, this application provides a computer device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the above-described knowledge graph-based on-site audit assistance method.

[0055] Fourthly, this application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-described knowledge graph-based on-site audit assistance method.

[0056] According to the specific embodiments provided in this application, this application has the following technical effects.

[0057] This application provides a knowledge graph-based on-site audit assistance method, device, equipment, and medium. This application models the audit system, automatically obtaining the corresponding audit model based on the audit project without requiring auditors to query it. Furthermore, it automatically obtains suspicious data based on the audit model, requiring auditors to conduct on-site verification of the suspicious data, thus reducing the workload of on-site verification. It further automatically generates audit records based on the knowledge graph and automatically generates audit working papers based on knowledge and experience, eliminating the need for auditors to write them. This application improves the automation level of audit work, thereby improving audit efficiency. Attached Figure Description

[0058] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0059] Figure 1 This is a flowchart illustrating a knowledge graph-based on-site auditing assistance method provided in one embodiment of this application.

[0060] Figure 2 This is a schematic diagram of the structure of a computer device provided in an embodiment of this application. Detailed Implementation

[0061] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0062] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0063] In one exemplary embodiment, a knowledge graph-based on-site audit assistance method is provided, such as... Figure 1 As shown, it includes the following steps 101-106.

[0064] Step 101: Determine the audit model corresponding to the audit project; the audit model is created based on the audit system corresponding to the audit project.

[0065] Step 102: Based on the audit model, review the audit data corresponding to the audit project to identify suspicious data;

[0066] Step 103: Use on-site verification to determine the evidence photos, verification results, and evidence collection records for each suspicious data point;

[0067] Step 104: Establish an evidence collection form based on the suspicious data that are verified to be true, along with the evidence collection photos and records.

[0068] Step 105: Using a knowledge graph, analyze the text data in the evidence collection table to generate audit records; the knowledge graph stores different audit standards and different historical audit records; the text data includes suspicious data and its evidence collection photos and records;

[0069] Step 106: Based on the audit records and experiential knowledge, automatically generate audit working papers; the experiential knowledge includes: an audit question bank and a case bank.

[0070] Implementing steps 101-106 above can improve the automation level of auditing work, thereby improving auditing efficiency.

[0071] Furthermore, traditional digital auditing methods cannot achieve full coverage of the project process and cannot effectively audit incomplete data. Incomplete business data access can prevent some audit procedures from being executed, ultimately affecting the quality of the project audit. Many audit tasks still need to be performed on-site, resulting in lower efficiency and cost, and failing to meet the audit supervision system requirements of new audit standards. To overcome these shortcomings, data access needs to be optimized technically, including the following three optimization methods:

[0072] The first is to establish unified data access standards and specifications, such as data format standardization, to formulate unified data format specifications to ensure that data from different sources can be successfully integrated, and to define standardized data interfaces and protocols to facilitate data exchange between different systems.

[0073] The second is to integrate multi-source heterogeneous data by using ETL tools to extract data from multiple data sources, clean and transform it, and then load it into the target database; build a data platform to achieve centralized management and sharing of data from different sources; and resolve semantic differences between different data sources through data mapping and semantic transformation.

[0074] The third is to establish a data supplementation and correction mechanism, formulate a process for supplementing missing data, promptly supplement key data, and use algorithms to automatically identify and correct common data errors.

[0075] For example, embodiments of this application use data imputation techniques to fill in missing data, enabling more comprehensive analysis of incomplete data. Examples include mean imputation (replacing missing values ​​with the mean of the dataset) and regression imputation (predicting missing values ​​using linear or nonlinear regression models). In addition, pattern recognition algorithms are applied. These algorithms help identify features and trends in the data, allowing for effective analysis even when the data is incomplete. Cluster analysis (hierarchical clustering) is used to discover natural groupings of the data; anomaly detection identifies potential outlier data points for special handling during imputation or analysis.

[0076] In another exemplary embodiment, step 101 above, which selects the corresponding audit model based on the scope of the audit project when the audit project is established, can be replaced by steps 201 and 202.

[0077] Step 201: Determine the audit scope. First, based on the company's specific needs and the audit project's objectives, determine the specific scope and key areas of the audit. For example, the audit scope may include financial statements, internal controls, compliance checks, etc.

[0078] Step 202: Select the appropriate audit model based on the audit scope. Based on the determined audit scope, select a suitable audit model from the pre-set audit model library. The audit model library can contain different types of audit models, such as financial audit models, compliance audit models, and operational audit models.

[0079] The following uses a financial audit model as an example to illustrate the audit model, which specifically includes:

[0080] I. Business Objectives of the Audit Model: Whether the use of cash complies with regulations and whether the amount used exceeds the limit.

[0081] II. Legal and regulatory basis (i.e., auditing system): Southern Power Grid regulations.

[0082] III. Business Data Reconciliation Relationships:

[0083] 1. Compare the payment details data (payment data from the company to employees) with the employee roster in the human resources system;

[0084] 2. Export payment details for personnel whose payment summaries do not include fields such as "travel, salary, bonus, allowance, welfare, heating".

[0085] IV. Income and Expenditure Identifier: This identifier exists in the underlying database of the financial control system.

[0086] 5. The employer receives the money: The employee pays the employer.

[0087] VI. Input Data:

[0088] 1. Final payment result table;

[0089] 2. Basic personnel roster;

[0090] 3. Bank transaction records.

[0091] VII. Output Data: Municipal Power Supply Company, County Power Supply Company, Power Supply Station, Basic Personnel Information, Summary, Income and Expenditure Identifier, Payment Amount, Payment Date.

[0092] In another exemplary embodiment, step 102 above generates suspicious data based on the selected audit model. This suspicious data is the focus of further verification and evidence collection.

[0093] The following section uses a selected compliance audit model (e.g., the power factor assessment standard audit model) to analyze relevant data of the enterprise and generate potentially problematic data.

[0094] I. Model Business Objective: To detect whether the power factor assessment standards stipulated in the "Electricity Supply Business Rules" are correctly implemented.

[0095] II. Legal and regulatory basis: China Southern Power Grid regulations.

[0096] III. Business Data Reconciliation Relationships:

[0097] Industrial users with high-voltage power supply of 160 kVA or above who do not comply with the power factor standard of 0.90 are considered incorrect users.

[0098] Industrial users with a power factor of 160 kVA (kW) and below, and other industrial users with a power factor greater than 100 kVA, are not subject to the power factor standard of 0.85 and are therefore considered incorrect users.

[0099] Non-industrial users with a power factor of 0.85 or higher are not subject to the power factor standard and are therefore considered incorrect users.

[0100] This applies to agricultural users and wholesale users with a capacity of 100 kVA (kW) and above. It does not adhere to the power factor standard of 0.80 and is therefore considered an incorrect user.

[0101] Large industrial users who are not designated as wholesale users directly managed by the power grid do not comply with the power factor standard of 0.85 and are therefore considered incorrect users.

[0102] IV. Required Data: Data Name: Customer Information Table; Key Data Fields: User ID, Username, Electricity Category, Electricity Capacity, Power Factor Standard.

[0103] V. Output Data: Data Name: Output Table of Users Who Did Not Correctly Implement Power Factor Assessment Standards

[0104] VI. Key data fields: User ID, Username, Transformer Capacity, Power Factor Standard, Electricity Consumption Category.

[0105] The company's business data is stored in the company's data center. The audit model runs in the company's data center. After the model runs, the suspicious data generated will be pushed to our database for storage.

[0106] In another exemplary embodiment, the above step 103, which involves viewing suspicious data via a mobile app and conducting on-site verification, and taking photos as evidence for any confirmed problems, can be replaced by the following steps 301-303.

[0107] Step 301: View suspicious data in real time via mobile app. Auditors can view suspicious data generated by the system in real time via mobile app to understand the specific content and location that needs to be verified.

[0108] Step 302: On-site verification of the suspicious data. Auditors will conduct on-site verification of the suspicious data provided by the app to confirm its accuracy and authenticity.

[0109] Step 303: Take photos as evidence for confirmed problems. For problems that have been verified, auditors use a mobile app to take photos as evidence, record detailed on-site conditions, and upload the photos to the system.

[0110] In another exemplary embodiment, steps 104 and 105 above will create an evidence collection form for the true doubts, and automatically generate audit records by relying on knowledge (such as implementation plans, audit notices, question databases, system databases, case databases, etc.) extraction and fusion, natural language processing and other technologies (such as similarity calculation, intent recognition, text classification, feature extraction, natural language generation models, etc.), specifically including the following steps 401 and 402.

[0111] Step 401: Create an evidence collection table for verified suspicious points. Summarize the verified suspicious point data, create an evidence collection table, and record the detailed information and evidence collection results for each suspicious point.

[0112] Step 402: Utilize natural language processing technologies (such as similarity calculation, intent recognition, text classification, feature extraction, and natural language generation models) to automatically generate audit records. The system analyzes the data in the evidence collection form using natural language processing technology, automatically generates detailed audit records, and saves the records to the system. This includes the following steps:

[0113] Step 402-1, Feature Extraction.

[0114] Feature extraction refers to extracting useful information from a knowledge topology to enable models to better understand and learn from the data. In the field of natural language processing, feature extraction typically refers to transforming text data into a representation that computers can process, such as word vectors. These features will be used in subsequent analysis and generation processes.

[0115] Step 402-2, Text Analysis and Understanding.

[0116] Similarity Calculation: The system can use natural language similarity calculation technology to compare the similarity between text data in the evidence collection form and target audit standards or historical audit records. This helps to identify potential anomalies or violations.

[0117] Intent recognition: Through intent recognition technology, the system can analyze user intent or behavioral purpose in text data. For example, during the auditing process, the system can identify intents such as requests, inquiries, and statements in text data, thereby more accurately understanding the text content.

[0118] Text Classification: The system can classify text data in the evidence collection form into predefined categories, such as compliant, non-compliant, and suspected issues. This helps to quickly identify content that requires priority attention. For example, the evidence collection form can be automatically categorized through text analysis, primarily through feature extraction, converting the text into feature vectors representing the frequency and importance of words. Alternatively, classification algorithms (SVM) can be used to automatically classify data using labeled training data.

[0119] Step 402-3: Generate audit records.

[0120] Information extraction: Extracting structured information from text data, such as time, location, people, and events. This information will serve as the foundation for audit logs.

[0121] Text generation: Utilizing a natural language generation model, detailed audit records are automatically generated based on extracted structured information and preset templates. These records should include key information such as the audit target, audit time, audit content, and audit results.

[0122] Record saving: Save the generated audit records to the system for later review and analysis.

[0123] The following section uses any text data 'i' from the evidence collection table as an example to explain in detail the process of generating audit records.

[0124] Calculate the similarity between text data i and each audit standard and each historical audit record in the knowledge graph to determine the behavioral classification result of text data i, which is compliant, non-compliant, or suspected problem; i = 1, 2, ..., I, where I is the number of text data in the evidence table; use an intent recognition model to identify the intent type of text data i; the intent type is a request, inquiry, or statement; use a natural language recognition model to extract structured information from text data i; the structured information includes: time, location, people, and events; based on the behavioral classification result, intent type, and structured information of text data i, use a preset template to generate the audit record of text data i.

[0125] The process of calculating the similarity between text data i and various audit standards and historical audit records in the knowledge graph to determine the behavior classification result of text data i includes: identifying the audit standard with the highest similarity to text data i in the knowledge graph as the target audit standard; determining whether text data i meets the target audit standard to obtain the first audit result of text data i; identifying the historical audit record with the highest similarity to text data i in the knowledge graph as the target audit record; determining the audit result in the target audit record as the second audit result of text data i; when both the first audit result and the second audit result are compliant, the behavior classification result is determined to be compliant; when both the first audit result and the second audit result are non-compliant, the behavior classification result is determined to be non-compliant; when one of the first audit result and the second audit result is compliant and the other is non-compliant, the behavior classification result is determined to be a suspected problem.

[0126] In another exemplary embodiment, step 105 above relies on knowledge (such as implementation plans, audit notices, question banks, system banks, case banks, etc.) extraction and fusion, knowledge graph technology, to extract and fuse relevant knowledge base information to generate a knowledge graph, including the following steps 501-505.

[0127] Step 501: Knowledge Extraction.

[0128] Data preprocessing:

[0129] Deduplication: Using hash functions or unique identifiers to detect and delete duplicate data records.

[0130] Data cleaning includes removing invalid characters, correcting spelling errors, and standardizing data format. For example, regular expressions can be used to match and replace specific character patterns.

[0131] Standardization: Converting data into a unified standard format to facilitate subsequent processing and analysis.

[0132] Step 502: Entity recognition.

[0133] Use Conditional Random Fields (CRF), Hidden Markov Models (HMM), or deep learning models (such as BERT) to identify entities in text.

[0134] The accuracy of entity recognition can be evaluated by calculating precision, recall, and F1 score.

[0135] Precision = Number of correctly identified entities / Total number of identified entities.

[0136] Recall rate = Number of correctly identified entities / Number of actual entities.

[0137] F1 score = 2 * (precision * recall) / (precision + recall).

[0138] Step 503: Relation extraction.

[0139] Rule-based methods: Matching and extracting relationships based on a predefined rule base.

[0140] Statistical methods: using machine learning models (such as Support Vector Machine (SVM), Naive Bayes (NB), etc.) to predict relationships.

[0141] Deep learning methods utilize neural network models (such as convolutional neural networks (CNNs) and recurrent neural networks (RNNs) to automatically learn relational features.

[0142] Step 504: Knowledge Integration.

[0143] Entity alignment:

[0144] Entity alignment across knowledge sources is a key step in the fusion process.

[0145] Similarity calculations (such as cosine similarity, Jaccard similarity, etc.) can be used to compare entities in different knowledge sources.

[0146] Cosine similarity = (A·B) / (||A||·||B||), where A and B are vectorized entities, and ||A|| and ||B|| are the magnitudes of A and B, respectively.

[0147] Entities are aligned based on a similarity threshold.

[0148] Relationship integration:

[0149] For aligned entities, it is necessary to integrate the relationships between them.

[0150] If the same relationship exists in different knowledge sources, they are merged; if conflicting relationships exist, conflict detection and resolution are required.

[0151] Step 505: Knowledge Graph Construction.

[0152] Graph databases (such as Neo4j) can be used to store knowledge graphs. Nodes in a knowledge graph represent entities, and edges represent relationships. Graph algorithms (such as PageRank, shortest path algorithms, etc.) can be used to analyze the structure and characteristics of the graph.

[0153] In another exemplary embodiment, step 106 above integrates experience and knowledge such as audit question bank and case bank to assist in intelligent writing of audit working papers, specifically including the following steps 601-602.

[0154] Step 601: Integrate experience and knowledge from the audit question bank and case library. The system integrates the experience and knowledge from the audit question bank and case library with the current audit records to provide reference and support for the writing of audit working papers. Specifically, this includes:

[0155] Step 601-1: Standardize the audit question bank.

[0156] Establish a structured audit question database, including: question name, project type, business area where the question occurred, stage at which the question occurred, question characterization, legal basis for characterization, liability definition, and analysis of the causes of the question.

[0157] Step 601-1: Standardize the case library.

[0158] Establish a structured case library, including: case name, project type, auditee, business area, problem description, and amount of loss or waste.

[0159] Step 601-3: Similarity matching and association.

[0160] Text similarity calculation: Using natural language processing technology, calculate the similarity between the current audit record and entries in the question and case databases.

[0161] Association rule mining: Through data mining techniques, potential associations between audit records and the issue and case databases are discovered.

[0162] Step 601-4: Extraction and recommendation of experiential knowledge.

[0163] Key information extraction: Extract key information from the matched entries, such as common problems, solutions, and risk points.

[0164] Intelligent recommendation system: Based on the extracted information, it provides intelligent recommendations for the writing of audit working papers, including possible audit findings and suggested actions.

[0165] Step 602: Assisted Intelligent Writing of Audit Working Papers. Based on integrated experience and current audit records, the system automatically generates audit working papers, improving the accuracy and completeness of audit reports.

[0166] Compared with existing technologies, this invention provides a knowledge graph-based on-site audit assistance method, which has the following beneficial effects:

[0167] 1. Improve audit efficiency: By selecting the appropriate audit model based on the audit scope when establishing an audit project, suspicious data is automatically generated, reducing the time and workload of manual screening; at the same time, suspicious data can be viewed in real time through a mobile APP, and on-site verification can be carried out. For problems that are indeed found, photos can be taken as evidence, simplifying the on-site audit process.

[0168] 2. Improve audit accuracy: Relying on knowledge extraction and fusion (such as implementation plans, audit notices, issue databases, policy databases, case databases, etc.) and natural language processing technologies (such as similarity calculation, intent recognition, text classification, feature extraction, natural language generation models, etc.), audit records are automatically generated, reducing errors and subjective biases in manual recording; at the same time, it integrates experiential knowledge such as audit issue databases and case databases to assist in the intelligent writing of audit working papers, thereby improving the accuracy and completeness of audit content.

[0169] 3. Supports rapid retrieval of large-scale institutional databases: By constructing and applying an audit knowledge graph, rapid and accurate retrieval of a vast institutional database is achieved, reducing the time and effort auditors spend on manual searches and improving the convenience and efficiency of audit work; at the same time, the system automatically recommends relevant rules and regulations, as well as historical issues and doubts, helping auditors quickly locate the information they need.

[0170] 4. Full-process coverage and incomplete data processing: This invention can achieve full-process coverage of the audit project, avoiding the shortcomings of traditional digital audit methods in terms of full-process coverage; when business data access is incomplete, the system can effectively process incomplete data to ensure the smooth execution of the audit procedure and improve the quality of project audit.

[0171] 5. High integration: The system modules of this invention and the coordination between them achieve efficient functional integration, meet the actual needs of auditors, and optimize the functional modules and the integration between modules of the audit information management platform.

[0172] In summary, this application, through innovative technical means and system design, significantly improves the efficiency and accuracy of on-site audit evidence collection, providing strong support for modern auditing work.

[0173] Based on the same inventive concept, this application also provides a knowledge graph-based on-site audit assistance device for implementing the knowledge graph-based on-site audit assistance method described above. The solution provided by this device is similar to the solution described in the above method. Therefore, the specific limitations of one or more knowledge graph-based on-site audit assistance device embodiments provided below can be found in the limitations of the knowledge graph-based on-site audit assistance method described above, and will not be repeated here.

[0174] In one exemplary embodiment, a knowledge graph-based on-site auditing assistance device is provided, comprising:

[0175] The audit model determination module is used to determine the audit model corresponding to the audit project; the audit model is created based on the audit system corresponding to the audit project.

[0176] The suspicious data identification module is used to review the audit data corresponding to the audit project based on the audit model and identify suspicious data.

[0177] The evidence collection module is used to determine the evidence photos, verification results, and evidence collection records for each suspicious data point through on-site verification.

[0178] The evidence collection form creation module is used to create an evidence collection form based on the suspicious data that has been verified as true, as well as the evidence collection photos and records.

[0179] The audit record generation module is used to analyze the text data in the evidence collection table using a knowledge graph to generate audit records; the knowledge graph stores different audit standards and different historical audit records; the text data includes suspicious data and its evidence collection photos and records.

[0180] The audit working paper generation module is used to automatically generate audit working papers based on the audit records and experiential knowledge; the experiential knowledge includes an audit question bank and a case bank.

[0181] In one exemplary embodiment, a computer device is provided, which may be a server or a terminal, and its internal structure diagram may be as follows. Figure 2 As shown, this computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operating system and computer programs stored in the non-volatile storage media to run. The I / O interfaces are used for exchanging information between the processor and external devices. The communication interface is used for communicating with external terminals via a network connection. When the computer program is executed by the processor, it implements a knowledge graph-based on-site auditing assistance method.

[0182] Those skilled in the art will understand that Figure 2 The structures shown are merely block diagrams of some structures related to the present application and do not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than shown in the figures, or combine certain components, or have different component arrangements. In an exemplary embodiment, a computer device is provided, including a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method embodiments.

[0183] In one exemplary embodiment, a computer-readable storage medium is provided storing a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0184] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties.

[0185] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM).

[0186] The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0187] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above 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.

[0188] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A knowledge graph-based on-site auditing assistance method, characterized in that, include: Determine the audit model corresponding to the audit project; The audit model is created based on the audit system corresponding to the audit project; Based on the audit model, the audit data corresponding to the audit project is reviewed to identify suspicious data; On-site verification was used to determine the evidence photos, verification results, and evidence collection records for each suspicious data point. An evidence collection form is established based on the verified suspicious data, along with the evidence photos and records. A knowledge graph is used to analyze the text data in the evidence collection table and generate audit records; the knowledge graph stores different audit standards and different historical audit records; the text data includes suspicious data and its evidence collection photos and records. Based on the audit records and experiential knowledge, audit working papers are automatically generated. The experiential knowledge includes: an audit question bank and a case bank; Using knowledge graphs, we analyze the textual data in the evidence collection form to generate audit records, specifically including: Calculate the similarity between text data i and each audit standard and each historical audit record in the knowledge graph to determine the behavioral classification result of text data i, wherein the behavioral classification result is compliant, non-compliant or suspected problem; i = 1, 2, ..., I, where I is the number of text data in the evidence table; Using an intent recognition model, the intent type of text data i is identified; the intent type is a request, an inquiry, or a statement. Structured information is extracted from text data i using a natural language recognition model; the structured information includes: time, location, people, and events; Based on the behavior classification results, intent type, and structured information of text data i, an audit record for text data i is generated using a preset template; Calculate the similarity between text data i and each audit standard and each historical audit record in the knowledge graph to determine the behavioral classification result of text data i, specifically including: The audit standard with the highest similarity to text data i in the knowledge graph is identified as the target audit standard. Determine whether text data i meets the target audit criteria, and obtain the first audit result for text data i; Identify the historical audit record in the knowledge graph that has the highest similarity to text data i, and use it as the target audit record; The audit results in the target audit record are determined as the second audit result of text data i; When both the first audit result and the second audit result are compliant, the behavior classification result is determined to be compliant; When both the first audit result and the second audit result are non-compliant, the behavior classification result is determined to be non-compliant; When one of the first audit result and the second audit result is compliant and the other is non-compliant, the behavior classification result is determined to be a suspected problem.

2. The knowledge graph-based on-site audit assistance method according to claim 1, characterized in that, The knowledge graph is constructed in the following manner; Obtain a database set; the sources of the knowledge text data in the database set include: audit notices, question databases, policy databases, and case databases; Preprocess the knowledge text data in the database set; Entity recognition and relation extraction are performed on each preprocessed knowledge text data to obtain an entity relation set; Merge the relationships of the same entity from different sources in the entity relationship set to obtain the entity relationship set after relationship merging. A knowledge graph is constructed based on the set of entity relationships.

3. The knowledge graph-based on-site audit assistance method according to claim 2, characterized in that, Preprocessing includes: Deduplication, data cleaning, and standardization; The deduplication method is as follows: determine that two pieces of knowledge text data whose absolute value of the difference between hash function values ​​in the database set is less than a hash threshold are duplicate data, and delete one of them; The data cleaning method is as follows: remove invalid characters from each knowledge text data and correct spelling errors in each knowledge text data.

4. The knowledge graph-based on-site audit assistance method according to claim 2, characterized in that, Merging relationships between the same entity from different sources within an entity relationship set yields a merged entity relationship set, specifically including: Calculate the cosine similarity between two entities from different sources within a set of entity relations; Determine whether two entities with a cosine similarity greater than a similarity threshold have the same relationship to obtain the first judgment result; If the first judgment result is yes, then merge the two entities and their relationship with a cosine similarity greater than the similarity threshold; If the first judgment result is negative, then determine whether the relationship between two entities with a cosine similarity greater than the first similarity threshold is conflicting, and obtain the second judgment result; If the second judgment result is negative, then retain the two entities and their relationship with a cosine similarity greater than the first similarity threshold. If the second judgment result is yes, delete the two entities and their relationship whose cosine similarity is greater than the first similarity threshold.

5. The knowledge graph-based on-site audit assistance method according to claim 1, characterized in that, Based on the audit records and experiential knowledge, audit working papers are automatically generated, including: Using natural language processing technology, audit question records in the audit question database with a similarity greater than a second similarity threshold are identified as first reference audit question records, and case records in the case database with a similarity greater than a third similarity threshold are identified as first reference case records. Using data mining techniques, audit issue records in the audit issue database that have potential connections with the audit record are identified as second reference audit issue records, and case records in the case database that have potential connections with the audit record are identified as second reference case records. Key information is extracted from each first reference audit issue record, each second reference audit issue record, each first case record, and each second case record; the key information includes: problem discovery, solutions, and risk points.

6. A knowledge graph-based on-site auditing auxiliary device, characterized in that, The knowledge graph-based on-site audit assistance device applies the knowledge graph-based on-site audit assistance method according to any one of claims 1-5, wherein the knowledge graph-based on-site audit assistance device comprises: The audit model determination module is used to determine the audit model corresponding to the audit project; the audit model is created based on the audit system corresponding to the audit project. The suspicious data identification module is used to review the audit data corresponding to the audit project based on the audit model and identify suspicious data. The evidence collection module is used to determine the evidence photos, verification results, and evidence collection records for each suspicious data point through on-site verification. The evidence collection form creation module is used to create an evidence collection form based on the suspicious data that has been verified as true, along with the evidence collection photos and records. The audit record generation module is used to analyze the text data in the evidence collection table using a knowledge graph to generate audit records; the knowledge graph stores different audit standards and different historical audit records; the text data includes suspicious data and its evidence collection photos and evidence collection records. The audit working paper generation module is used to automatically generate audit working papers based on the audit records and experiential knowledge; the experiential knowledge includes an audit question bank and a case bank.

7. A computer device, comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that the processor executes the computer program to implement the knowledge graph-based on-site auditing assistance method according to any one of claims 1-5.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the computer program implements the knowledge graph-based on-site auditing assistance method as described in any one of claims 1-5.

Images