Work report classification auditing method and device

Abstract

The invention discloses a work report classification auditing method and device, relates to the technical field of work report classification auditing, and mainly aims to improve the efficiency and accuracy of work report classification auditing. According to the main technical scheme, the method comprises the steps of generating work report data according to report content of a to-be-audited work report and a target report category to which the work report is divided; determining target report knowledge data matched with the work report, wherein the target report knowledge data supports distinguishing different report categories from at least one dimension; generating cue words based on the target report knowledge data, wherein the cue words are used for guiding an auditing model to combine the target report knowledge data and the work report data to audit whether the work report is correctly divided into the target report category; and calling an auditing model according to the cue word, performing classified auditing on the work report based on the target report knowledge data and the work report data, and generating an auditing result for indicating whether the work report is correctly classified into the target report category or not.

Description

Technical Field

[0001] This application relates to the field of work report classification and review technology, and in particular to a work report classification and review method and apparatus. Background Technology

[0002] In organizations such as businesses, employee work reports are crucial for recording, summarizing, and reporting on work performance. Organizations have numerous report categories, necessitating performance evaluations, cost analyses, and service quality monitoring based on these categories. However, employee reports may be mistakenly categorized incorrectly. If subsequent monitoring is conducted based on this incorrect category, it can easily lead to monitoring irregularities and potentially disrupt the overall workflow. Therefore, verifying the correct report category for each work report is a critical step in effectively preventing monitoring irregularities.

[0003] Traditional methods for classifying and reviewing work reports primarily rely on manual processing. Manual review depends on dedicated auditors reading each report individually and judging, based on experience, whether a report has been correctly categorized. This method is limited by the auditors' subjective understanding, professional level, and workload, resulting in low efficiency and accuracy in work report classification and review. This not only makes it difficult to handle the large-scale classification and review needs of work reports but also frequently leads to review errors.

[0004] Therefore, how to improve the efficiency and accuracy of work report classification and review has become an urgent problem to be solved. Summary of the Invention

[0005] This application proposes a method and apparatus for classifying and reviewing work reports, with the main purpose of improving the efficiency and accuracy of classifying and reviewing work reports.

[0006] To achieve the above objectives, this application mainly provides the following technical solutions:

[0007] Firstly, this application provides a method for classifying and reviewing work reports. The method provided in this embodiment may include at least the following steps: generating work report data based on the report content of the work report to be reviewed and the target report category to which the work report is classified; determining target report knowledge data matching the work report, wherein the target report knowledge data supports distinguishing different report categories from at least one dimension; generating prompt words based on the target report knowledge data, wherein the prompt words guide the review model to review whether the work report is correctly classified into the target report category by combining the target report knowledge data and the work report data; and, based on the prompt words, calling the review model to classify and review the work report according to the target report knowledge data and the work report data, generating a review result indicating whether the work report is correctly classified into the target report category.

[0008] Secondly, this application provides a work report classification and review device, which in this embodiment may include at least:

[0009] The first generation module is used to generate work report data based on the report content of the work report to be reviewed and the target report category to which the work report is classified.

[0010] A determination module is used to determine the target report knowledge data that matches the work report, wherein the target report knowledge data supports distinguishing different report categories from at least one dimension;

[0011] The second generation module is used to generate prompt words based on the target report knowledge data. The prompt words are used to guide the review model to review whether the work report is correctly classified into the target report category by combining the target report knowledge data and the work report data.

[0012] The review module is used to invoke the review model based on the prompt words, and to classify and review the work report based on the target report knowledge data and the work report data, and generate a review result indicating whether the work report has been correctly classified into the target report category.

[0013] Thirdly, this application provides a computer-readable storage medium including a stored program, wherein the program, when running, controls the device where the storage medium is located to execute the work report classification and review method of the first aspect.

[0014] Fourthly, this application provides an electronic device comprising: a memory for storing a program; and a processor coupled to the memory for running the program to perform the work report classification and review method of the first aspect.

[0015] Fifthly, this application provides a computer program product, which includes: a computer program / computer executable instructions, wherein the computer program / computer executable instructions are a work report classification and review method of the first aspect.

[0016] The work report classification and review method and apparatus provided in this application, when there is a work report to be reviewed, firstly generates work report data based on the report content and the target report category to which the work report is classified. Then, it determines the target report knowledge data that matches the work report and supports distinguishing different report categories from at least one dimension. Next, it generates prompt words based on the target report knowledge data. These prompt words guide the review model to review whether the work report has been correctly classified into the target report category, combining the target report knowledge data and the work report data. Finally, based on the prompt words, the review model is invoked to classify and review the work report based on the target report knowledge data and the work report data, generating a review result indicating whether the work report has been correctly classified into the target report category. Therefore, the solution provided in this application can generate work report data based on the report content and the report category to which the work report is classified, and dynamically generate prompt words based on the work report data and the matching report knowledge data. Subsequently, the generated prompt words guide the review model to automatically review whether the work report has been correctly classified into the report category, combining the report knowledge data and the work report data. This not only significantly improves review efficiency by replacing manual review with a review model, but also enables precise review of the report categories by accurately analyzing the correctness of the report categories through dynamically generated prompts, thereby improving the accuracy of the review.

[0017] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, the following are specific embodiments of this application. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 A flowchart illustrating a work report classification and review method according to an embodiment of this application is shown;

[0020] Figure 2 This illustration shows a schematic diagram of a dimension traversal process provided in one embodiment of this application;

[0021] Figure 3 A schematic diagram of an audit model provided in one embodiment of this application is shown;

[0022] Figure 4 This illustration shows a structural schematic diagram of a work report classification and review device according to an embodiment of this application;

[0023] Figure 5 A schematic diagram of the structure of a work report classification and review device provided in another embodiment of this application is shown. Detailed Implementation

[0024] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0025] In organizations such as enterprises, work reports are often categorized into various types based on business functions. For example, software development companies often categorize work reports into pre-sales (e.g., covering customer needs coordination and solution writing), project-based (e.g., covering project progress and resource allocation), and after-sales (e.g., covering issue response and satisfaction feedback). Different report categories require different regulatory operations, such as performance evaluation, cost analysis, and service quality monitoring. If a work report is incorrectly categorized and subsequent regulatory operations are performed based on this incorrect category, it can easily lead to regulatory irregularities. Therefore, it is necessary to verify the correctness of the report category for each work report.

[0026] Based on research findings, work report data can be generated from the content of the work report to be reviewed and the report category it belongs to. Then, prompts are dynamically generated based on the work report data and matching report knowledge data. These prompts then guide the review model to automatically verify whether the work report has been correctly categorized, combining the report knowledge data with the work report data. This not only significantly improves review efficiency by replacing manual review with a review model, but also achieves precise review of the correctness of the work report's category through dynamically generated prompts, thus improving the accuracy of the review.

[0027] Based on the above findings, this application provides a specific technical solution for classifying and reviewing work reports, including: generating work report data based on the content of the work report to be reviewed and the target report category to which the work report is classified; determining the target report knowledge data that matches the work report, wherein the target report knowledge data supports distinguishing different report categories from at least one dimension; generating prompt words based on the target report knowledge data, wherein the prompt words are used to guide the review model to review whether the work report is correctly classified into the target report category by combining the target report knowledge data and the work report data; and calling the review model based on the prompt words to classify and review the work report based on the target report knowledge data and the work report data, and generating a review result indicating whether the work report is correctly classified into the target report category.

[0028] The work report classification and review technology solution provided in this embodiment can be applied to the work report review system of any organization to improve the efficiency and accuracy of the work report review system in classifying and reviewing work reports for the organization. This embodiment does not limit the industry of the organization; it can be flexibly selected based on business needs. For example, industries may include, but are not limited to: software development, cybersecurity, and finance.

[0029] Based on the above-mentioned technical solution for classifying and reviewing work reports, this embodiment specifically provides a method and apparatus for classifying and reviewing work reports. The method and apparatus for classifying and reviewing work reports provided in this embodiment will be described in detail below.

[0030] This application provides a method for classifying and reviewing work reports, such as... Figure 1 As shown, the work report classification and review method provided in this embodiment may include at least the following steps 101 to 104.

[0031] 101. Generate work report data based on the content of the work report to be reviewed and the target report category to which the work report is classified.

[0032] In some embodiments, organizations such as enterprises may have work reports that have been written, categorized, and are awaiting submission to the appropriate terminals for subsequent performance evaluation, cost analysis, and service quality monitoring, etc., according to established procedures. Because employee errors (e.g., selecting the wrong report category when writing the report) may result in the work report being categorized incorrectly, it is necessary to identify the work report as a pending review report before sending it to the terminal for monitoring. Methods for identifying pending review reports can include at least the following methods A1 and A2.

[0033] Method A1 involves monitoring the target storage location for newly stored work reports. This location receives and stores work reports submitted by employee terminals within the organization and awaiting forwarding to those terminals for oversight. If a newly stored work report is detected, it is identified as a work report awaiting review. This ensures that any newly submitted work report can be promptly classified and reviewed as such.

[0034] Method A2, upon receiving an audit instruction, designates the work report specified in the audit instruction as the work report to be audited. This allows for more flexible designation of work reports to be audited, improving the focus and efficiency of the audit.

[0035] The above methods A1 and A2 can be flexibly selected for use based on business needs, and this embodiment does not limit this.

[0036] In some embodiments, after identifying the work report to be reviewed, a step is performed to generate work report data based on the report content of the work report to be reviewed and the target report category to which the work report is classified, in order to provide the core data of "work report data" for work report classification review. The process of generating work report data based on the report content of the work report to be reviewed and the target report category to which the work report is classified may include at least the following steps 101A to 101B.

[0037] 101A. Modify any unusual words in the work report to standard terminology used in the target industry to which the work report pertains.

[0038] The target industry of a work report is the basis for identifying anomalous words in its content; therefore, it is necessary to determine the target industry of the work report. Methods for determining the target industry of a work report can include at least one of the following: First, determine the employee attribute information of the employee who submitted the work report; then determine the industry indicated by the employee attribute information as the target industry matching the work report. For example, if the employee attribute information of the employee who submitted the work report is: position: backend development engineer; department: software R&D department, then based on this employee attribute information, the target industry matching the work report is determined to be the software development industry. Second, extract the field values ​​corresponding to preset fields from the work report; these preset fields are industry-related fields; then determine the industry described by the field values ​​as the target industry matching the work report.

[0039] Employees may use terms that don't conform to industry standards when writing work reports due to their own writing habits, or they may make typos. These terms will be considered abnormal. Therefore, abnormal terms can include, but are not limited to, at least one of the following: terms that do not conform to the target industry's terminology, or terms containing typos. The presence of abnormal terms reduces the accuracy of work report classification and review. Therefore, it is necessary to modify abnormal terms in work report content to replace them with standard terminology from the target industry to standardize the work report content.

[0040] The process of modifying abnormal words in a work report to standard terminology for the target industry can include: determining a target standard terminology library applicable to the target industry from a pre-defined standard terminology library. Each pre-defined standard terminology library has an applicable industry, and the library records the mapping relationship between abnormal words and the standard terminology of the applicable industry. Then, based on the mapping relationship recorded in the target standard terminology library, abnormal words in the work report content are traversed, and these abnormal words are modified to standard terminology for the target industry. It should be noted that the mapping relationship in the standard terminology library can include at least one of the following: a mapping relationship between words that do not meet industry terminology requirements and the industry's standard terminology; a mapping relationship between words containing misspellings and the industry's standard terminology that does not contain misspellings.

[0041] For example, the target standard terminology library includes: mapping relationships between words that do not meet industry terminology requirements and standard industry terms {"product installation": "product deployment", "POC": "test verification", "product demo": "product demonstration"}, and mapping relationships between words containing typos and standard industry terms that do not contain typos {"deployment": "deployment", "fault": "malfunction"}. After traversal, the abnormal words in the work report content include "deployment" and "product installation". Subsequently, "deployment" in the work report content is modified to "deployment", and "product installation" is modified to "product deployment", thereby completing the standardization of the work report content. The code for this process can be written based on a specific programming language. For example, if the programming language is Python, the implementation code may include, but is not limited to:

[0042]

[0043] 101B. Combine the category labels corresponding to the target report category with the modified work report content to generate work report data.

[0044] Work reports are associated with category labels, which indicate the target report category to which the work report is classified. For example, if a work report is classified into the target report category "Project Category", then it is associated with the category label "Project Class".

[0045] The method for generating work report data by concatenating the category label corresponding to the target report category and the modified work report content can include either of the following two approaches: One is to directly concatenate strings. First, store the category label and the work report content in two variables, then combine them using string concatenation operators (such as ":" or "-"). For example, if the category label is "Project" and the work report content is "Core module development completed this week," then the concatenated work report data would be "Project: Core module development completed this week." This method is suitable for scenarios that do not require complex formatting, can be quickly implemented using basic string operations, and facilitates the subsequent unified storage or display of the concatenated results. Another approach is to first define a dictionary with fixed keys to store relevant information. For example, create a dictionary `report = {"category": "Project", "content": "Core module development completed this week"}`, where the "category" key corresponds to the category label "Project", and the "content" key corresponds to the work report content "Core module development completed this week". Then, use the f-string formatting function to concatenate the data according to a preset template, i.e., execute `result = f"{report['category']}-{report['content']}"`, and finally get the concatenated result "Project - Core module development completed this week". This method clarifies the data storage structure and allows for flexible adjustment of the concatenation format, making it convenient for batch processing of similar data.

[0046] 102. Determine the target report knowledge data to match the work report. The target report knowledge data should support the differentiation of different report categories from at least one dimension.

[0047] In some embodiments, the core purpose of determining the target report knowledge data that matches the work report is to establish a connection between the work report and the known target report knowledge data, so as to accurately review the report category of the work report through the known target report knowledge data.

[0048] The process of determining the matching report knowledge data for a work report may include at least the following steps 102A to 102C.

[0049] 102A. Based on the target information, determine the target industry to which the work report belongs.

[0050] Different industries have different applicable report knowledge data. Therefore, in order to more accurately review the report category of a work report, it is necessary to obtain the target information of the work report and, based on the target information, determine the target industry to which the work report belongs. In this way, the target report knowledge data matching the work report can be obtained from the report knowledge data corresponding to the target industry.

[0051] The target information may include, but is not limited to, at least one of the following: First, employee attribute data corresponding to the employee who wrote the work report, which reflects the industry to which the work report belongs. For example, if the employee attribute information is: position is backend development engineer, department is software R&D department, this employee attribute information indicates that the target industry of the work report is the software development industry. Second, the field value corresponding to the first field extracted from the work report, where the first field is a field related to the industry to which the work report belongs. The first field can be determined based on business needs, and this embodiment does not limit it. For example, the first field includes department name, project name, and position name.

[0052] After identifying the target information, the next step is to determine the target industry to which the work report belongs. This step may include: querying the correspondence between sample information and industries; if there is sample information that matches the target information in the correspondence, the industry corresponding to the matching sample information is determined as the target industry to which the work report belongs; if there is no sample information that matches the target information in the correspondence, a prompt is issued based on the target information, so that the reviewers can manually verify the target industry to which the work report belongs based on the prompt.

[0053] 102B. Select the target report knowledge base corresponding to the target industry from the preset report knowledge base. Each preset report knowledge base has a corresponding industry.

[0054] To improve the efficiency of acquiring target report knowledge data for matching work reports, multiple industry-specific report knowledge bases are pre-defined, each containing report knowledge data for its respective industry. It should be noted that each industry's report knowledge base will be continuously updated and improved over time to enrich and refine its content. Furthermore, if a new industry is discovered, a corresponding report knowledge base will be built for that new industry.

[0055] In some embodiments, the report knowledge base corresponding to each industry may include at least one type, and this embodiment does not limit this. For example, the report knowledge base corresponding to each industry may include, but is not limited to, the following four types: job definition library, job activity definition library, context clue library, and thesaurus.

[0056] The work definition library records the standardized definitions for each report category and the work activities included in each category. Specifically, the work definition library uses fixed structured fields (e.g., report category name, standardized definition, work activity name, etc.) to define each report category. The work activity definition library records the definition descriptions and characteristics of the work activities included in each report category for each report category recorded in the work definition library. The context clue library records at least one dimension of report knowledge data, i.e., clue data, for each report category recorded in the work definition library. Dimensions can include, but are not limited to, at least one of time, purpose, and scenario dimensions. Time-dimensional clue data clarifies the execution cycle or nodes of the work. Purpose-dimensional clue data points to the core objective of the work. Scenario-dimensional clue data describes the execution scenario and dependencies of the work, assisting in determining the report category to which the work belongs, and further improving the accuracy of report category identification when combined with work activity characteristics. The thesaurus records synonymous expressions of the data in the aforementioned three knowledge bases.

[0057] For example, the work definition library records the following for the report category "Pre-sales Category": the report category name is "Pre-sales Category," the standardized definition is "Technical verification conducted during the sales cycle to facilitate a transaction," and the work activity name is "Equipment Deployment and Testing, Product Demonstration." The work activity definition library records the following for the work activity "Product Fault Handling" within the report category "Pre-sales Category": the work activity name is "Product Fault Handling," the definition description is "Pre-sales Exclusive," and the work activity characteristics are "occurring during the sales stage, with the purpose of eliminating customer technical concerns." The context clue library records the following for the report category "Pre-sales Category": the time-dimensional clue data is "Sales Stage, Bidding Period"; the purpose-dimensional clue data is "Facilitating a Transaction, Winning a Customer"; and the scenario-dimensional clue data is "Demonstration, Testing, POC." The thesaurus records corresponding synonyms (e.g., synonyms) for "Fault Handling" in the work activity name "Product Fault Handling" (e.g., synonyms) such as "Problem Handling, Troubleshooting, Anomaly Handling, Filling Problem Handling." The code for this type of report knowledge base can be written in a specific programming language. For example, if the programming language is Python, the implementation code may include, but is not limited to:

[0058]

[0059] 102C. Obtain target report knowledge data matching the work report from the target report knowledge base.

[0060] After identifying the target industry, the system first filters out the target report knowledge base corresponding to that industry from the pre-set report knowledge base. Then, it specifically determines the target report knowledge data to match the work report from the target report knowledge base. This approach avoids interference from irrelevant information in non-target industry knowledge bases, preventing knowledge mismatches due to industry differences. It also ensures that the matched target report knowledge data closely aligns with the target industry, thereby improving the accuracy of the target report knowledge data and its suitability for the work report, providing more accurate support for the classification and review of work reports.

[0061] The target report knowledge base typically includes report knowledge data corresponding to various report categories. Therefore, in order to more effectively obtain target report knowledge data that matches work reports, the implementation method for obtaining target report knowledge data that matches work reports from the target report knowledge base can at least include any one of the following schemes B1 and B2.

[0062] Method B1 involves retrieving the target report knowledge data that matches the work report from the target report knowledge base, corresponding to the target report category to which the work report is classified.

[0063] This method is suitable for scenarios where only the accuracy of the work report's classification into the target report category needs to be verified, without specifying the correct category. This reduces the amount of knowledge data required for the target report, thereby improving the efficiency of verifying the correct classification.

[0064] Method B2 involves extracting the target report knowledge data that matches the work report from the target report knowledge base for each report category.

[0065] This method is applied to scenarios where work reports are reviewed to ensure they have been correctly categorized into the target report category, and where, if a work report is found to have been incorrectly categorized, the correct report category can be provided quickly without manual intervention.

[0066] The above methods B1 and B2 can be used flexibly based on business needs, and this embodiment does not limit the choice of either method.

[0067] 103. Generate prompt words based on target report knowledge data. The prompt words are used to guide the audit model to combine target report knowledge data with work report data to audit whether the work report has been correctly classified into the target report category.

[0068] After determining the target report knowledge data to match the work report, prompt words are generated based on this data. These prompt words guide the review model to verify whether the work report has been correctly categorized into the target report category, combining the target report knowledge data with the work report data. The core purpose of this approach is to ensure that the prompt words accurately anchor the actual content and knowledge association of the work report, effectively avoiding classification bias or dimensional omissions that may occur with generic prompt words. In this method, the prompt words can rely on the matched target report knowledge data to clearly define the classification criteria, thereby significantly improving the accuracy of the review results and reducing misjudgments caused by information ambiguity in the review model, making the review process more efficient and more aligned with actual needs.

[0069] The process of generating prompt words based on report knowledge data may include the following steps: obtaining a prompt word template matching the work report, the prompt word template including at least one second field to be filled; extracting the target field string corresponding to each second field from the report knowledge data; using each target string as the value of the corresponding second field to fill the prompt template and generate the prompt word.

[0070] To improve the effectiveness of prompt generation, multiple prompt templates corresponding to different industries can be preset. After determining the industry to which the work report belongs, the preset prompt templates corresponding to that industry will be used to obtain the matching prompt template for the work report.

[0071] For example, the obtained prompt word template matching the work report is as follows:

[0072]

[0073] It should be noted that the content to be analyzed in the prompt word module is used to indicate the work report data that needs to be filled in the work report.

[0074] After obtaining the prompt word template matching the work report, the first step is to break down the structure of the prompt word template, identify all the second fields to be filled (e.g., reporter, work activity name, etc.), and clarify the definition and expected data requirements of each second field. Next, the knowledge data is comprehensively sorted and filtered. Based on the data requirements of each second field, the specific data that accurately corresponds to each second field is extracted from the target report knowledge data through keyword matching, semantic association analysis, etc., i.e., the target field strings. These target field strings are then processed into standardized and complete strings. Finally, according to the field order and position of the prompt word template, each extracted and processed target field string is filled into the corresponding second field. After filling, the overall content is checked to ensure that the value of each second field is accurate and without omissions or errors, and finally, the prompt words are generated.

[0075] 104. Based on the prompt words, the audit model is invoked to classify and audit the work reports according to the target report knowledge data and work report data, and to generate audit results indicating whether the work reports have been correctly classified into the target report category.

[0076] In some embodiments, the target report knowledge data covers at least one first report category and includes first report knowledge data for at least one dimension corresponding to each first report category. The dimensions may include, but are not limited to, at least one of the following: time dimension, purpose dimension, and scenario dimension. Cue words indicate the priority order of at least one dimension.

[0077] Based on this, the process of calling the review model according to the prompt words to classify and review the work report based on the target report knowledge data and the work report data, and generating a review result indicating whether the work report has been correctly classified into the target report category, may include the following steps: calling the review model according to the prompt words and traversing the dimensions one by one in priority order; if the review model analyzes that there is a second report category in the first report category under the current dimension, then the second report category is compared with the target report category as the report category to be confirmed, and a review result indicating whether the work report has been correctly classified into the target report category is generated, where the work report data successfully matches the category clue data corresponding to the second report category under the current dimension, and does not successfully match the category clue data corresponding to other report categories in the first report category besides the second report category under the current dimension; if the review model does not analyze that there is a second report category in the first report category under the current dimension, then it continues to traverse the next dimension until it analyzes that there is a second report category in the first report category, or until all dimensions have been traversed.

[0078] Specifically, based on the prompt words, the audit model is invoked and traversed to each dimension in priority order. Then, for the currently traversed dimension, the following steps 104A to 104C are executed.

[0079] 104A. Determine whether the audit model has analyzed whether a second report category exists in the first report category under the current dimension. If yes, proceed to step 104B; otherwise, proceed to step 104C.

[0080] The review model is a pre-trained model with the ability to classify and review work reports. Its specific type can be flexibly selected based on business needs, and this embodiment does not limit it.

[0081] If a second report category is determined to exist, it means that the work report data has successfully matched the category clue data corresponding to the second report category in the current dimension, and has not successfully matched the category clue data corresponding to other report categories in the first report category besides the second report category in the current dimension. The correct report category of the work report is likely the second report category, so proceed to step 104B.

[0082] If it is determined that there is no second report category, it means that the correct report category of the work report cannot be determined yet, and further judgment is required. Therefore, step 104C is executed.

[0083] 104B. Compare the second report category as the report category to be confirmed with the target report category to generate an audit result indicating whether the work report has been correctly classified into the target report category.

[0084] The specific process of comparing the report category to be confirmed with the target report category to generate an audit result indicating whether the work report has been correctly classified into the target report category can be as follows: The comparison involves checking if the report category to be confirmed is the same as the target report category. If the comparison shows the report category to be confirmed is the same as the target report category, it means the work report has been correctly classified into the target report category, and thus an audit result indicating that the work report has been correctly classified into the target report category is generated. If the comparison shows the report category to be confirmed is different from the target report category, it means the work report has been incorrectly classified into the target report category, and an audit result indicating that the work report needs to be correctly classified into the report category to be confirmed is generated. Based on such audit results, the work report can be promptly corrected to be classified into the correct report category.

[0085] 104C. Determine if the current dimension is the last dimension in the priority order. If yes, then determine that all dimensions have been traversed. If not, continue traversing the next dimension.

[0086] If the audit model does not identify a second report category within the first report category under the current dimension, it determines whether the current dimension is the last dimension in the priority order to decide whether to terminate the traversal process. If the current dimension is determined to be the last dimension in the priority order, the traversal process terminates. If the current dimension is not determined to be the last dimension in the priority order, it means that there are still dimensions that have not been traversed, and the traversal continues to the next dimension.

[0087] In some embodiments, exemplarily, the following are examples Figure 2 The diagram shown illustrates the traversal process. Figure 2The dimensions involved include time, purpose, and scenario, with the priority order being: time before purpose, and purpose before scenario. The review model is invoked based on the prompts, traversing each dimension in priority order.

[0088] The first dimension traversed is the time dimension, and the category clue data corresponding to the time dimension is used for stage identification. If the audit model analyzes that the work report data successfully matches the category clue data "Sales Stage" corresponding to the pre-sales category in the time dimension, then the pre-sales category is determined as the second report category. If the audit model analyzes that the work report data successfully matches the category clue data "Implementation Stage" corresponding to the project category in the time dimension, then the project category is determined as the second report category. If the audit model analyzes that the work report data successfully matches the category clue data "Operation Stage" corresponding to the after-sales category in the time dimension, then the after-sales category is determined as the second report category. If the audit model analyzes that the work report data fails to match the category clue data corresponding to the pre-sales, project, and after-sales categories in the time dimension, and therefore cannot be identified, then the next dimension, i.e., the target dimension, is traversed according to priority.

[0089] The next dimension traversed is the target dimension. If the audit model analyzes the work report data and successfully matches the pre-sales category's corresponding category lead data "facilitating the transaction" in the target dimension, then the pre-sales category is determined as the second report category. If the audit model analyzes the work report data and successfully matches the project category's corresponding category lead data "ensuring delivery" in the target dimension, then the project category is determined as the second report category. If the audit model analyzes the work report data and successfully matches the after-sales category's corresponding category lead data "ensuring operation" in the target dimension, then the after-sales category is determined as the second report category. If the audit model analyzes the work report data and fails to successfully match the pre-sales, project, and after-sales categories' corresponding category lead data in the target dimension (i.e., it cannot be identified), then the next dimension, the scenario dimension, is traversed according to priority.

[0090] The current dimension being iterated over is the scenario dimension. If the audit model successfully matches the work report data with the category lead data "including demo / POC" corresponding to the pre-sales category in the scenario dimension, then the pre-sales category is determined as the second report category. If the audit model successfully matches the work report data with the category lead data "including deployment / cutover" corresponding to the project category in the scenario dimension, then the project category is determined as the second report category. If the audit model successfully matches the work report data with the category lead data "including production / fault" corresponding to the after-sales category in the scenario dimension, then the after-sales category is determined as the second report category. If the audit model fails to match the work report data with the category lead data corresponding to the pre-sales, project, and after-sales categories in the scenario dimension (i.e., it cannot be identified), then the scenario dimension is determined as the last dimension according to priority, and the traversal process is terminated. After terminating the traversal process, in order to achieve work report classification audit, the default boundary of the audit model is used to predict the report category of the work report based on the work report data. The work report is classified and audited by comparing the target report category with the predicted report category.

[0091] In some embodiments, the target report knowledge data also includes second report knowledge data related to the basic work attributes corresponding to each first report category, with prompt words indicating that the second report knowledge data has a higher priority than the first report knowledge data. Based on this, considering that work reports can be correctly classified and reviewed based on the second report knowledge data, there is no need to use the first report knowledge data for classification and review, thereby reducing the complexity of the analysis. Therefore, the work report classification and review method provided in this embodiment may further include the following steps 104D to 104E.

[0092] 104D. Based on the prompt words, call the audit model to determine whether the target report knowledge data includes the target data; if it does, execute the above steps to call the audit model based on the prompt words and traverse the dimensions one by one in priority order; if it does not, execute step 104E.

[0093] The target data is used to indicate the report category for which the work report cannot be reviewed based on the second report knowledge data. In other words, for a work report containing the target data, if the second report knowledge data is insufficient to correctly review the report category, the first report knowledge data is required for review. In some embodiments, the target data is data that appears and is used in work reports corresponding to two or more report categories, and it can be flexibly set according to actual needs. For example, the target data is the field string "Problem Handling," which may appear in work reports of the pre-sales, project, and after-sales categories.

[0094] If the audit model is invoked based on the prompt words and determines that the target report knowledge data includes the target data, it indicates that it is difficult to accurately audit the work report classification using the second report knowledge data related to the basic work attributes. Therefore, the above steps are performed to invoke the audit model based on the prompt words and traverse the dimensions one by one in priority order to use the second report knowledge data for work report classification audit.

[0095] If the audit model determines that the target report knowledge data does not include the target data based on the prompt words, it means that the work report classification can be accurately audited by using the second report knowledge data related to the basic work attributes, and there is no need to use the first report knowledge data. Therefore, in order to reduce the audit complexity, step 104E is executed to audit the work report classification based on the second report knowledge data.

[0096] For example, such as Figure 2 As shown, the target data is the field string "Problem Handling". First, determine if the string "Problem Handling" is present. If it is, execute the above steps, and based on the prompt words, call the review model to traverse the dimensions one by one in priority order. Figure 2 The first dimension traversed is the time dimension; if it is determined that it is not included, then the basic attribute analysis of the job is performed. Figure 2 The analysis of the basic attributes of the work in the text specifically refers to step 104E.

[0097] 104E. Based on the prompt words, the audit model is invoked to classify and audit the work report according to the knowledge data of the second report and the work report data.

[0098] The second report knowledge data includes at least one type of job attribute clue data, and the prompt words indicate the dependency order of the at least one type of job attribute clue data. For example, job attribute clue data may include, but is not limited to, at least one of the following types: job definition type and job activity definition type. The job activity definition type depends on the job definition type and is located after the job definition type in the dependency order. The second report knowledge data corresponding to the job activity definition type is used to define the job activities involved in the job defined by the second report data corresponding to the job definition.

[0099] Based on this, the process of classifying and reviewing work reports according to the prompt words and the review model based on the second report knowledge data and work report data can include the following steps: The review model is invoked according to the prompt words and iterates through the types one by one in the dependency order; if it is determined that the work report data successfully matches the work attribute clue data in the knowledge data to be matched corresponding to the current type, then the work attribute clue data of the next type corresponding to the first report category to which the successfully matched work attribute clue data belongs is taken as the knowledge data to be matched and continues to be matched with the work report data; if the work report data successfully matches the work attribute clue data in the knowledge data to be matched of the last type, then the first report category to which the successfully matched work attribute clue data belongs is determined as the third report category; the third report category is compared with the target report category as the report category to be confirmed, and a review result is generated to indicate whether the work report has been correctly classified into the target report category.

[0100] Specifically, based on the prompt words, the audit model is invoked to traverse the types one by one in the order of dependencies. Then, for the currently traversed type, the following steps 104E1 to 104E4 are executed.

[0101] 104E1. Determine whether the work report data has been successfully matched with the work attribute clue data in the knowledge data to be matched for the current type. If yes, proceed to step 104E2. If no, terminate the type traversal process.

[0102] The knowledge data to be matched for the current type falls into two categories: First, if the current type is the first type in the dependency order, the knowledge data to be matched is all the job attribute clue data corresponding to the current type; second, if the current type is not the first type in the dependency order, the knowledge data to be matched is the job attribute clue data of the report category to which the successfully matched job attribute clue data in the previous type belongs in the current type.

[0103] If the work report data is determined to be successfully matched with the work attribute clue data in the knowledge data to be matched in the current type, it means that the report category to which the successfully matched work attribute clue data belongs is likely to be the correct report category of the work report. Therefore, in order to further verify, it is necessary to continue to execute step 104E2.

[0104] If the work report data fails to find a matching work attribute clue in the knowledge data to be matched for the current type, it means that the analysis of the work report cannot be determined based on the knowledge data of the second report, and the subsequent traversal process is meaningless. Therefore, the type traversal process is terminated.

[0105] 104E2. The next type of work attribute clue data corresponding to the first report category to which the successfully matched work attribute clue data belongs is taken as the knowledge data to be matched and continues to be matched with the work report data. This method of traversing types in dependency order can automatically connect to the next type of knowledge data to be matched after a successful match, ensuring the logical consistency of knowledge association and avoiding matching gaps.

[0106] 104E3. If the work report data successfully matches the work attribute clue data in the last type of knowledge data to be matched, then the first report category to which the successfully matched work attribute clue data belongs is determined as the third report category.

[0107] If the work report data successfully matches the work attribute clue data in the last type of knowledge data to be matched, it means that the correct category of the work report is the first report category to which the successfully matched work attribute clue data belongs. At this time, in order to verify the correctness of the target report category to which the work report is currently classified, the first report category to which the successfully matched work attribute clue data belongs is determined as the third report category.

[0108] 104E4. Compare the third report category as the report category to be confirmed with the target report category to generate an audit result indicating whether the work report has been correctly classified into the target report category.

[0109] The detailed explanation of step 104E4 is basically the same as that of step 104B above, so it will not be repeated here.

[0110] For example, the job attribute clue data includes at least one of the following types: job definition type and job activity definition type. The dependency order is that the job definition type precedes the job activity definition type. The target report knowledge data includes job attribute clue data related to the job definition type corresponding to the pre-sales category: the standardized definition is "technical verification carried out during the sales cycle to facilitate the transaction", and the job activity name is "equipment deployment and testing, product demonstration, product troubleshooting". The target report knowledge data also includes job attribute clue data related to the job activity definition type corresponding to the pre-sales category: the job activity name is "product troubleshooting", the definition description is "pre-sales exclusive", and the job activity characteristic is "occurring in the sales stage, with the purpose of eliminating customer technical doubts". First, the job activity definition type is traversed according to the dependency order, and it is determined that the job report data successfully matches the job attribute clue data "product troubleshooting" in the knowledge data to be matched corresponding to the job activity definition type. Then, the job attribute clue data of the next type "job activity definition type" corresponding to the first report category "pre-sales category" to which the successfully matched job attribute clue data "product troubleshooting" belongs is taken as the knowledge data to be matched and continues to be matched with the job report data. After matching, it was determined that the work report data successfully matched the work attribute clue data "occurred in the sales stage" in the knowledge data to be matched corresponding to the work activity definition type, and the pre-sales category was determined to be the third report category.

[0111] In some embodiments, considering that there are different synonymous expressions for work attribute clue data, in order to avoid the inability to perform work report classification and review due to the application of synonyms and other data, the work report classification and review method provided in this embodiment may further include the following steps: if it is determined that the work report data has not been successfully matched with work attribute clue data in the knowledge data to be matched corresponding to the current type, then a thesaurus corresponding to the current type is determined, the thesaurus being used to record synonymous expressions of the work attribute clue data corresponding to the current type; if it is determined that the work report data has been successfully matched with synonymous expressions in the thesaurus corresponding to the current type, then the work report data has been successfully matched with work attribute clue data in the knowledge data to be matched corresponding to the current type.

[0112] When it is determined that no work attribute clue data can be successfully matched in the corresponding knowledge data for the current type of work report data, a thesaurus is introduced to break down matching barriers caused by differences in expression. The thesaurus covers synonyms of the report knowledge data, including cases where "the expression is different but the core meaning is the same" in the effective matching range, avoiding knowledge omissions caused by differences in word usage habits and expression methods, and significantly improving the comprehensiveness and accuracy of matching.

[0113] In some embodiments, when reviewing reports based on the first report knowledge data, it is considered that a second report category may not be determined even after traversing all dimensions. Similarly, when reviewing reports based on the second report knowledge data, a third report category may not be determined even after traversing all types. Therefore, to ensure that the review can be completed, the work report classification review method provided in this embodiment may further include the following steps 104F to 104H.

[0114] 104F If the second report category is still not determined after traversing all dimensions, and / or if the third report category is still not determined after traversing all types, then the audit model is invoked to match the corresponding sample work report for the work report.

[0115] If a second report category cannot be determined after traversing all dimensions, and / or if a third report category cannot be determined after traversing all types, it indicates that the target report knowledge data based on the acquired work report matching cannot be used to review the work report classification. In this case, to ensure the sequential review of work report classification, the review model is invoked to match the work report with a corresponding sample work report. The report category of the work report is determined by the similarity between the work report and the sample work report. The sample work report is a work report that is known to have been correctly classified into the corresponding report category.

[0116] Methods for determining the sample work report to match the work report can include the following two: First, pre-set multiple sample work reports corresponding to different industries. Determine the target industry to which the work report belongs, and then identify the sample work report corresponding to that target industry as the matching sample work report. Second, such as... Figure 3 As shown, Figure 3 A schematic diagram of the review model is shown. The review model includes a first feature extractor and a sample matching engine. First, the first feature extractor extracts features from the work report data. Then, the sample matching engine matches the extracted features with the features of sample work reports. If a match is successful, the successfully matched sample work report is obtained as the work report matching sample work report. If a match fails, the work report is provided to the expert review terminal for review. If the expert review terminal confirms the correct report category, the analyzer generates a decision to use the work report as a new sample work report. Then, a second feature extractor (which may or may not reuse the first feature extractor) extracts features for generating the new sample work report. New sample work report features are then generated based on these features and incrementally added to the sample database used to store existing sample work reports, continuously enriching and improving the sample database.

[0117] 104G. Based on the similarity between the work report data and the sample work report data corresponding to the sample work report, the target sample work report is determined.

[0118] like Figure 3 As shown, the review module also includes a rule applicator. The rule applicator is invoked to calculate the similarity between the work report data and the sample work report data corresponding to the sample work report using a preset similarity calculation method. The similarity calculation method can be flexibly selected based on business needs, and this embodiment does not limit it.

[0119] Specifically, if it is determined that there are sample work report data with a similarity threshold, then the sample work report corresponding to the sample work report data that reaches the similarity threshold will be identified as the target sample work report.

[0120] Specifically, if it is determined that there is no sample work report data with a similarity threshold, it means that there is no sample work report data corresponding to the work report data. At this time, it is difficult to determine the report type of the work report, and at least one of the following operations should be performed: First, issue a prompt that the work report cannot be reviewed, so that the reviewer can manually classify and review the work report based on the prompt; Second, in order to continuously improve the work report review capability, if the expert review terminal provides feedback that the work report has the correct report category, add the work report as a new sample work report increment to the original sample work report, so that when encountering other work reports that are the same or similar, the review can be carried out based on the new sample work report.

[0121] 104H: Compare the report category of the target sample work report as the report category to be confirmed with the target report category to generate an audit result indicating whether the work report has been correctly classified into the target report category. The detailed explanation of step 104H is basically the same as that of step 104B above, so it will not be repeated here.

[0122] In some embodiments, after receiving the audit results, if the audit results indicate that the work report has been correctly classified into the target report category, it means that the work report can be normally monitored, and therefore the work report can be directly sent to the corresponding terminal for monitoring. If the audit results indicate that the work report has not been correctly classified into the target report category, it means that the work report cannot be normally monitored, and the report category of the work report needs to be corrected. In this case, the work report along with the audit results can be sent to each audit terminal so that the audit terminal can correct the category of the work report, or so that the auditor corresponding to the audit terminal can review the report category of the work report.

[0123] The work report classification and review method provided in this application, when a work report to be reviewed exists, firstly generates work report data based on the report content and the target report category to which the work report is classified. Then, it determines the target report knowledge data that matches the work report and supports distinguishing different report categories from at least one dimension. Next, it generates prompt words based on the target report knowledge data. These prompt words guide the review model to review whether the work report has been correctly classified into the target report category, combining the target report knowledge data and the work report data. Finally, based on the prompt words, the review model is invoked to classify and review the work report based on the target report knowledge data and the work report data, generating a review result indicating whether the work report has been correctly classified into the target report category. Therefore, the solution provided in this application can generate work report data based on the report content and the report category to which the work report is classified, and dynamically generate prompt words based on the work report data and the matching report knowledge data. Subsequently, the generated prompt words guide the review model to automatically review whether the work report has been correctly classified into the report category, combining the report knowledge data and the work report data. This not only significantly improves review efficiency by replacing manual review with a review model, but also enables precise review of the report categories by accurately analyzing the correctness of the report categories through dynamically generated prompts, thereby improving the accuracy of the review.

[0124] Furthermore, one embodiment of this application also provides a work report classification and review device, such as... Figure 4 As shown, the work report classification and review device provided in this embodiment may include at least:

[0125] The first generation module 21 is used to generate work report data based on the report content of the work report to be reviewed and the target report category to which the work report is classified.

[0126] The determination module 22 is used to determine the target report knowledge data that matches the work report, wherein the target report knowledge data supports distinguishing different report categories from at least one dimension;

[0127] The second generation module 23 is used to generate prompt words based on the target report knowledge data. The prompt words are used to guide the review model to review whether the work report is correctly classified into the target report category by combining the target report knowledge data and the work report data.

[0128] The review module 24 is used to call the review model based on the prompt words to classify and review the work report according to the target report knowledge data and the work report data, and generate a review result indicating whether the work report has been correctly classified into the target report category.

[0129] The work report classification and review device provided in this application, when a work report to be reviewed exists, first generates work report data based on the report content and the target report category to which the work report is classified. Then, it determines the target report knowledge data that matches the work report and supports distinguishing different report categories from at least one dimension. Next, it generates prompt words based on the target report knowledge data. These prompt words guide the review model to review whether the work report has been correctly classified into the target report category, combining the target report knowledge data and the work report data. Finally, based on the prompt words, the review model is invoked to classify and review the work report based on the target report knowledge data and the work report data, generating a review result indicating whether the work report has been correctly classified into the target report category. Therefore, the solution provided in this application can generate work report data based on the report content and the report category to which the work report is classified, and dynamically generate prompt words based on the work report data and the matching report knowledge data. Subsequently, the generated prompt words guide the review model to automatically review whether the work report has been correctly classified into the report category, combining the report knowledge data and the work report data. This not only significantly improves review efficiency by replacing manual review with a review model, but also enables precise review of the report categories by accurately analyzing the correctness of the report categories through dynamically generated prompts, thereby improving the accuracy of the review.

[0130] In some embodiments of this application, such as Figure 5 As shown, the target report knowledge data covers at least one first report category and includes first report knowledge data of at least one dimension corresponding to each first report category, and the prompt words indicate the priority order of the at least one dimension; then the review module 24 may include:

[0131] Traversal unit 241 is used to call the review model according to the prompt words and traverse the dimensions one by one in the priority order;

[0132] The first review unit 242 is used to compare the second report category as the report category to be confirmed with the target report category if the review model analyzes that there is a second report category in the first report category under the current dimension, and generate a review result to indicate whether the work report is correctly classified into the target report category. The work report data successfully matches the category clue data corresponding to the second report category under the current dimension, and does not successfully match the category clue data corresponding to other report categories in the first report category besides the second report category under the current dimension.

[0133] The loop unit 243 is used to continue traversing the next dimension if the audit model does not find that there is a second report category in the first report category under the current dimension, until the analysis shows that there is a second report category in the first report category, or until all dimensions have been traversed.

[0134] In some embodiments of this application, such as Figure 5 As shown, the first review unit 242 is specifically used to generate a review result indicating that the work report is correctly classified into the target report category if the comparison shows that the report category to be confirmed is the same as the target report category; if the comparison shows that the report category to be confirmed is different from the target report category, it generates a review result indicating that the work report is not correctly classified into the target report category, and indicates in the review result that the work report needs to be correctly classified into the report category to be confirmed.

[0135] In some embodiments of this application, such as Figure 5 As shown, the target report knowledge data also includes second report knowledge data related to the basic work attributes corresponding to each of the first report categories. The prompt word indicates that the second report knowledge data has a higher usage priority than the first report knowledge data. Therefore, the review module 24 may further include:

[0136] The judgment unit 244 is used to call the review model based on the prompt word to determine whether the target report knowledge data includes target data, where the target data is used to indicate the report category that cannot be reviewed based on the second report knowledge data; if it is included, the traversal unit 241 is triggered to execute the step of calling the review model based on the prompt word and traversing the dimensions one by one according to the priority order; if it is not included, the second review unit 245 is triggered to call the review model based on the prompt word and classify and review the work report based on the second report knowledge data and the work report data.

[0137] In some embodiments of this application, such as Figure 5As shown, the second report knowledge data includes at least one type of work attribute clue data, and the prompt word indicates the dependency order of the at least one type of work attribute clue data. The second review unit 245 is specifically used to call the review model according to the prompt word and traverse the types one by one according to the dependency order. If it is determined that the work report data successfully matches work attribute clue data in the matching knowledge data corresponding to the current type, then the work attribute clue data of the next type corresponding to the first report category to which the successfully matched work attribute clue data belongs is taken as matching knowledge data and matched with the work report data again. If the work report data successfully matches work attribute clue data in the matching knowledge data of the last type, then the first report category to which the successfully matched work attribute clue data belongs is determined as the third report category. The third report category is compared with the target report category as the report category to be confirmed, generating a review result indicating whether the work report has been correctly classified into the target report category.

[0138] In some embodiments of this application, such as Figure 5 As shown, the second review unit 245 is specifically used to generate a review result indicating that the work report is correctly classified into the target report category if the comparison shows that the report category to be confirmed is the same as the target report category; if the comparison shows that the report category to be confirmed is different from the target report category, it generates a review result indicating that the work report is not correctly classified into the target report category, and indicates in the review result that the work report needs to be correctly classified into the report category to be confirmed.

[0139] In some embodiments of this application, such as Figure 5 As shown, the second review unit 245 can also be used to determine the thesaurus corresponding to the current type if it is determined that the work report data has not been successfully matched with work attribute clue data in the knowledge data to be matched corresponding to the current type. The thesaurus is used to record the synonym expression data of the work attribute clue data corresponding to the current type. If it is determined that the work report data has been successfully matched with the synonym expression data in the thesaurus corresponding to the current type, it is determined that the work report data has been successfully matched with work attribute clue data in the knowledge data to be matched corresponding to the current type.

[0140] In some embodiments of this application, such as Figure 5 As shown, the work attribute clue data involved in the second review unit 245 includes at least one of the following types: work definition type and work activity definition type. The second report knowledge data corresponding to the work activity definition type is used to define the work activities involved in the work defined by the second report data corresponding to the work definition.

[0141] In some embodiments of this application, such as Figure 5 As shown, the audit module 24 provided in this embodiment may further include: a third audit unit 246, used to call the audit model to match a corresponding sample work report for the work report if the loop unit 243 has traversed all dimensions and still has not determined the second report category, wherein the sample work report is a work report that is known to have been correctly classified into the corresponding report category; to determine the target sample work report based on the similarity between the work report data and the sample work report data corresponding to the sample work report; and to compare the report category of the target sample work report as the report category to be confirmed with the target report category to generate an audit result indicating whether the work report has been correctly classified into the target report category.

[0142] In some embodiments of this application, such as Figure 5 As shown, the third review unit 246 can also be used to, if it is determined that there is sample work report data with a similarity threshold, identify the sample work report corresponding to the sample work report data that reaches the similarity threshold as the target sample work report; if it is determined that there is no sample work report data with a similarity threshold, perform at least one of the following operations: issue a prompt that the work report cannot be reviewed, and / or, if the expert review terminal provides feedback that the work report has the correct report category, add the work report as a new sample work report increment to the original sample work report.

[0143] In some embodiments of this application, such as Figure 5 As shown, the third review unit 246 is specifically used to generate a review result indicating that the work report is correctly classified into the target report category if the comparison shows that the report category to be confirmed is the same as the target report category; if the comparison shows that the report category to be confirmed is different from the target report category, it generates a review result indicating that the work report is not correctly classified into the target report category, and indicates in the review result that the work report needs to be correctly classified into the report category to be confirmed.

[0144] In some embodiments of this application, such as Figure 5 As shown, the determining module 22 is specifically used to determine the target industry to which the work report belongs based on target information. The target information includes at least one of the following: employee attribute data corresponding to the employee who wrote the work report, and field values ​​corresponding to a first field extracted from the work report, wherein the first field is a field related to the industry to which the work report belongs; selecting a target report knowledge base corresponding to the target industry from a preset report knowledge base, wherein each preset report knowledge base has a corresponding industry; and obtaining target report knowledge data matching the work report from the target report knowledge base.

[0145] In some embodiments of this application, such as Figure 5 As shown, the first generation module 21 is specifically used to modify abnormal words in the work report content into standard terms of the target industry to which the work report belongs. The abnormal words include at least one of the following: words that do not conform to the terminology requirements of the target industry, words containing typos; and to concatenate the category label corresponding to the target report category and the modified work report content to generate work report data.

[0146] In some embodiments of this application, such as Figure 5 As shown, the second generation module 23 is specifically used to obtain the prompt word template matching the work report. The prompt word template includes at least one second field to be filled; extract the target field string corresponding to each second field from the report knowledge data; and use each target string as the value of the corresponding second field to fill the prompt template and generate the prompt word.

[0147] In some embodiments of this application, such as Figure 5 As shown, the determination module 22 involves at least one dimension, including at least one of the following: time dimension, purpose dimension, and scenario dimension.

[0148] For a detailed explanation of the operation of each functional module in the work report classification and review device provided in this application embodiment, please refer to the corresponding detailed explanation of the above-mentioned work report classification and review method embodiment, which will not be repeated here.

[0149] Furthermore, one embodiment of this application also provides a computer-readable storage medium, the storage medium including a stored program, wherein, when the program is executed, it controls the device where the storage medium is located to perform the above-described work report classification and review method.

[0150] Furthermore, one embodiment of this application also provides an electronic device, the electronic device comprising: a memory for storing a program; and a processor coupled to the memory for running the program to perform the above-described work report classification and review method.

[0151] Furthermore, one embodiment of this application also provides a computer program product, the computer program product comprising: a computer program / computer executable instructions, the computer program / computer executable to perform the above-described work report classification and review method.

[0152] In the above embodiments, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0153] It is understood that the relevant features in the above methods and apparatus can be referenced interchangeably. Furthermore, the terms "first," "second," etc., in the above embodiments are used to distinguish between embodiments and do not represent the superiority or inferiority of any particular embodiment.

[0154] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0155] The algorithms and displays provided herein are not inherently related to any particular computer, virtual system, or other device. Various general-purpose systems can also be used in conjunction with the teachings herein. The required structure for constructing such systems is apparent from the above description. Furthermore, this application is not directed to any particular programming language. It should be understood that the content of this application described herein can be implemented using various programming languages, and the above description of specific languages ​​is for the purpose of disclosing preferred embodiments of this application.

[0156] In addition, the memory may include non-permanent memory in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM, and the memory includes at least one memory chip.

[0157] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0158] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data cutover device to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data cutover device, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0159] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data cutover device to operate in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0160] These computer program instructions can also be loaded onto a computer or other programmable data cutover device to cause a series of operational steps to be performed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable device for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0161] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0162] Memory may include non-persistent memory in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, like read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0163] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0164] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0165] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0166] The above are merely embodiments of this application and are not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A work report classification review method characterized by, The method comprises: generating work report data according to the report content of the work report to be audited and the target report category into which the work report is classified; determining target report knowledge data matched by the work report, the target report knowledge data supporting distinguishing different report categories from at least one dimension; generating a prompt word based on the target report knowledge data, the prompt word being used to guide an audit model to combine the target report knowledge data and the work report data to audit whether the work report is correctly classified into the target report category; calling the audit model based on the prompt word to perform classified auditing on the work report based on the target report knowledge data and the work report data, and generating an audit result indicating whether the work report is correctly classified into the target report category.

2. The method of claim 1, wherein, The target report knowledge data covers at least one first report category and includes first report knowledge data of at least one dimension corresponding to each first report category, and the prompt word indicates a priority order of the at least one dimension; then, calling the audit model based on the prompt word to perform classified auditing on the work report based on the target report knowledge data and the work report data, and generating an audit result indicating whether the work report is correctly classified into the target report category, comprises: calling the audit model according to the prompt word to traverse dimensions one by one in the priority order; if the audit model analyzes that there is a second report category in the first report category under the current dimension, the second report category is taken as a report category to be confirmed and compared with the target report category, the work report data successfully matches category clue data corresponding to the second report category under the current dimension, and does not successfully match category clue data corresponding to other report categories in the first report category except the second report category under the current dimension, and the audit result indicating whether the work report is correctly classified into the target report category is generated; if the audit model does not analyze that there is a second report category in the first report category under the current dimension, the next dimension is continued to be traversed until a second report category in the first report category is analyzed or all dimensions are traversed.

3. The method of claim 2, wherein, The target report knowledge data further includes second report knowledge data corresponding to each first report category and related to work basic attributes, and the prompt word indicates that the use priority of the second report knowledge data is higher than that of the first report knowledge data; then, the method further comprises: calling the audit model based on the prompt word to determine whether the target report knowledge data includes target data, the target data being used to indicate that the report category of the work report cannot be audited based on the second report knowledge data; if yes, performing the step of calling the audit model according to the prompt word to traverse dimensions one by one in the priority order; if no, calling the audit model based on the prompt word to perform classified auditing on the work report based on the second report knowledge data and the work report data.

4. The method of claim 3, wherein, The second report knowledge data includes at least one type of work attribute clue data, and the prompt word indicates a dependency order of the at least one type of work attribute clue data, and the audit model is called according to the prompt word to perform classified auditing on the work report based on the second report knowledge data and the work report data, including: The audit model is called according to the prompt word to sequentially traverse the types according to the dependency order; If it is determined that the work report data successfully matches the work attribute clue data in the to-be-matched knowledge data corresponding to the current type, the work attribute clue data of a next type corresponding to a first report category to which the successfully matched work attribute clue data belongs is taken as to-be-matched knowledge data, and the matching with the work report data is continued; If the work report data successfully matches the work attribute clue data in the to-be-matched knowledge data of the last type, the first report category to which the successfully matched work attribute clue data belongs is determined as a third report category; The third report category is taken as a to-be-confirmed report category to be compared with the target report category, and an auditing result indicating whether the work report is correctly divided into the target report category is generated.

5. The method of claim 4, wherein, The method further includes: if it is determined that the work report data does not successfully match the work attribute clue data in the to-be-matched knowledge data corresponding to the current type, a synonym library corresponding to the current type is determined, the synonym library being used to record synonymous expression data of the work attribute clue data corresponding to the current type; if it is determined that the work report data successfully matches the synonymous expression data in the synonym library corresponding to the current type, it is determined that the work report data successfully matches the work attribute clue data in the to-be-matched knowledge data corresponding to the current type; and / or, The work attribute clue data includes at least one of the following types: a work definition type and a work activity definition type, and second report knowledge data corresponding to the work activity definition type is used to define work activities involved in work defined by second report data corresponding to the work definition.

6. The method of claim 2, wherein, The method further includes: If all dimensions are traversed and the second report category is not determined, an audit model is called to match a corresponding sample work report for the work report, the sample work report being a work report that is known to be correctly divided into a corresponding report category; A target sample work report is determined based on a similarity between the work report data and sample work report data corresponding to the sample work report; A report category of the target sample work report is taken as a to-be-confirmed report category to be compared with the target report category, and an auditing result indicating whether the work report is correctly divided into the target report category is generated.

7. The method of claim 6, wherein, The method further includes: If it is determined that there is sample work report data with a similarity reaching a similarity threshold, a sample work report corresponding to the sample work report data with the similarity reaching the similarity threshold is determined as the target sample work report. If it is determined that there is no sample work report data with a similarity reaching a similarity threshold, at least one of the following operations is performed: a prompt that the work report cannot be audited is issued, and / or, in the case that the expert audit terminal feeds back that the work report is correct in the report category, the work report is added to the original sample work report as a new sample work report increment.

8. The method according to any one of claims 2, 4 and 6, characterized in that, The to-be-confirmed report category is compared with the target report category, and an audit result indicating whether the work report is correctly classified into the target report category is generated, including: If the to-be-confirmed report category is compared with the target report category, the audit result indicating that the work report is correctly classified into the target report category is generated; If the to-be-confirmed report category is compared with the target report category, the audit result indicating that the work report is not correctly classified into the target report category is generated, and it is indicated in the audit result that the work report needs to be correctly classified into the to-be-confirmed report category.

9. The method according to any one of claims 1-7, characterized in that, The target report knowledge data matched by the work report is determined, including: determining a target industry to which the work report belongs based on target information, the target information including at least one of: employee attribute data corresponding to an employee who writes the work report, and a field value corresponding to a first field extracted from the work report, the first field being a field related to the industry to which the work report belongs; selecting a target report knowledge base corresponding to the target industry from a preset report knowledge base, each report knowledge base in the preset report knowledge base having a corresponding industry; and obtaining the target report knowledge data matched by the work report from the target report knowledge base. And / or, The work report data is generated according to the report content of the to-be-audited work report and the target report category to which the work report is classified, including: modifying an abnormal word in the work report content into a standard term of the target industry to which the work report belongs, the abnormal word including at least one of: a word that does not meet the language requirements of the target industry, and a word containing a wrong character; and splicing the category label corresponding to the target report category and the modified work report content to generate the work report data. And / or, The prompt word is generated based on the report knowledge data, including: obtaining a prompt word template matched by the work report, the prompt word template including at least one second field to be filled; extracting a target field string corresponding to each second field from the report knowledge data; and filling each target string as a value of the corresponding second field in the prompt template to generate the prompt word. And / or, The at least one dimension includes at least one of: a time dimension, a purpose dimension, and a scene dimension.

10. A work report classification review apparatus characterized by comprising: The device includes: The first generation module is configured to generate work report data according to the report content of the to-be-audited work report and the target report category to which the work report is classified. The determination module is configured to determine target report knowledge data matched by the work report, the target report knowledge data supporting the classification of different report categories from at least one dimension. The second generation module is configured to generate a prompt word based on the target report knowledge data, and the prompt word is used to guide an audit model to combine the target report knowledge data and the work report data to audit whether the work report is correctly classified into the target report category; The audit module is configured to invoke the audit model to perform classification auditing on the work report based on the target report knowledge data and the work report data according to the prompt word, and generate an audit result used to indicate whether the work report is correctly classified into the target report category.

11. A computer readable storage medium, characterized in that, The storage medium includes a stored program, wherein the program controls a device where the storage medium is located to perform the work report classification auditing method in any one of claims 1 to 9 when the program is running.

12. An electronic device, comprising: The electronic device includes a memory configured to store a program, and a processor coupled to the memory and configured to run the program to perform the work report classification auditing method in any one of claims 1 to 9.

13. A computer program product, characterised in that, The computer program product includes computer program / computer executable instructions configured to perform the work report classification auditing method in any one of claims 1 to 9.

Images