Report generation method and device, electronic equipment and storage medium

Abstract

Embodiments of the present application provide a report generation method and device, electronic equipment and storage medium. The method comprises: receiving a report topic; obtaining multi-source data, the multi-source data comprising structured data and unstructured data from heterogeneous data sources; performing retrieval enhancement generation processing on the multi-source data to obtain context information related to the report topic from the multi-source data, and generating a retrieval result according to the context information; and generating a target report according to the retrieval result. According to the above scheme, by using retrieval enhancement generation technology, the problem of information omission and subjective interference can be avoided compared with manual analysis, accurate context can be obtained, and the accuracy of the generated target report can be ensured based on accurate information, thereby improving the accuracy of the report.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, and in particular to a method, apparatus, electronic device, and storage medium for generating a report. Background Technology

[0002] Reports generated by analyzing raw data play a crucial role in various business operations, and the quality of these reports directly determines the accuracy of business decisions.

[0003] In related technologies, reports are generated by manually analyzing raw data.

[0004] However, manual analysis may contain omissions or subjective interference, leading to low accuracy in the report. Summary of the Invention

[0005] This application provides a method, apparatus, electronic device, and storage medium for generating reports, in order to improve the accuracy of reports.

[0006] In a first aspect, embodiments of this application provide a method for generating a report, comprising: receiving a report topic; acquiring multi-source data, the multi-source data including structured data and unstructured data from heterogeneous data sources; performing retrieval enhancement generation processing on the multi-source data to obtain contextual information related to the report topic from the multi-source data, and generating retrieval results based on the contextual information; and generating a target report based on the retrieval results.

[0007] In one possible implementation, a retrieval enhancement generation process is performed on the multi-source data to obtain contextual information related to the report topic from the multi-source data, and retrieval results are generated based on the contextual information. This includes: performing semantic similarity retrieval on the multi-source data based on the report topic to obtain a set of text fragments semantically related to the report topic from the multi-source data; inputting the set of text fragments as enhanced context into a pre-trained large language model to obtain the retrieval results, wherein the large language model is trained based on historical text.

[0008] In one possible implementation, generating a target report based on the search results includes: obtaining an initial report template, the initial report template including objective data fields; extracting standardized data corresponding to the objective data fields from the structured data using robotic process automation (RoLAD) technology; filling the standardized data into the positions corresponding to the objective data fields in the initial report template to obtain an intermediate report template; and generating the target report based on the search results and the intermediate report template.

[0009] In one possible implementation, obtaining the initial report template includes: receiving a service type identifier sent by the user terminal; obtaining a corresponding service rule from a preset service rule base based on the service type identifier; and dynamically generating the initial report template based on the service rule.

[0010] In one possible implementation, extracting standardized data corresponding to the objective data field from the structured data specifically includes: determining the data specification requirements corresponding to the objective data field based on the initial report template; determining the data extraction rules and source mapping relationship based on the data specification requirements; and extracting the standardized data corresponding to the objective data field from the structured data through an application programming interface based on the data extraction rules and source mapping relationship.

[0011] In one possible implementation, generating the target report based on the search results and the intermediate report template includes: combining the search results with the intermediate report template to obtain a combined report; performing content consistency verification and data rationality verification on the combined report to obtain a verification result, wherein the verification result is either verification passed or verification failed; and in response to the verification result being verification passed, formatting the combined report according to a preset report format to obtain the target report.

[0012] In one possible implementation, the search results are combined with the intermediate report template to obtain a combined report, including: inputting the search results into multiple vertical analysis models to obtain multiple intermediate analysis conclusions, wherein the multiple vertical analysis models are obtained by training models using sample data from different fields; fusing the multiple intermediate analysis conclusions through a preset decision fusion rule or model to obtain a target analysis conclusion; and filling the target analysis conclusion into the corresponding position of the intermediate report template to obtain the target report.

[0013] Secondly, embodiments of this application provide a report generation apparatus, comprising: a receiving module for receiving a report topic; an acquisition module for acquiring multi-source data, the multi-source data including structured data and unstructured data from heterogeneous data sources; a retrieval module for performing retrieval enhancement generation processing on the multi-source data to obtain contextual information related to the report topic from the multi-source data and generate retrieval results based on the contextual information; and a generation module for generating a target report based on the retrieval results.

[0014] In one possible implementation, the retrieval module is specifically configured to perform semantic similarity retrieval in the multi-source data based on the report topic, so as to obtain a set of text fragments semantically related to the report topic from the multi-source data; the retrieval module is further configured to use the set of text fragments as enhanced context and input it into a pre-trained large language model to obtain the retrieval results, wherein the large language model is trained based on historical text.

[0015] In one possible implementation, the generation module is specifically used to obtain an initial report template, the initial report template including objective data fields; the generation module is further used to extract standardized data corresponding to the objective data fields from the structured data using robotic process automation technology; the generation module is further used to fill the standardized data into the positions corresponding to the objective data fields in the initial report template to obtain an intermediate report template; the generation module is further used to generate the target report based on the search results and the intermediate report template.

[0016] In one possible implementation, the generation module is specifically configured to receive a service type identifier sent by the user terminal; the generation module is further configured to obtain a corresponding service rule from a preset service rule library based on the service type identifier; and the generation module is further configured to dynamically generate the initial report template based on the service rule.

[0017] In one possible implementation, the apparatus further includes: an extraction module, configured to determine the data specification requirements corresponding to the objective data field based on the initial report template; the extraction module is further configured to determine data extraction rules and source mapping relationships based on the data specification requirements; and the extraction module is further configured to extract standardized data corresponding to the objective data field from the structured data through an application programming interface based on the data extraction rules and source mapping relationships.

[0018] In one possible implementation, the apparatus further includes: an execution module, configured to combine the search results with the intermediate report template to obtain a combined report; the execution module is further configured to perform content consistency verification and data rationality verification on the combined report to obtain a verification result, wherein the verification result is either verification passed or verification failed; the execution module is further configured to, in response to the verification result being verification passed, format the combined report according to a preset report format to obtain the target report.

[0019] In one possible implementation, the execution module is specifically used to input the retrieval results into multiple vertical analysis models to obtain multiple intermediate analysis conclusions, wherein the multiple vertical analysis models are obtained by training models using sample data from different fields; the execution module is also specifically used to fuse the multiple intermediate analysis conclusions using preset decision fusion rules or models to obtain target analysis conclusions; the execution module is also specifically used to fill the target analysis conclusions into the corresponding positions of the intermediate report template to obtain the target report.

[0020] Thirdly, embodiments of this application provide a report generation device, including: a memory and a processor;

[0021] The memory stores computer-executed instructions;

[0022] The processor executes computer execution instructions stored in the memory, causing the processor to perform the first aspect and / or various possible implementations of the first aspect as described above.

[0023] Fourthly, embodiments of this application provide a non-volatile computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, are used to implement the first aspect and / or various possible implementations of the first aspect.

[0024] Fifthly, embodiments of this application provide a computer program product, including a computer program that, when executed by a processor, implements the first aspect and / or various possible implementations of the first aspect.

[0025] The report generation method, apparatus, electronic device, and storage medium provided in this application include: receiving a report topic; acquiring multi-source data, the multi-source data including structured and unstructured data from heterogeneous data sources; performing retrieval enhancement generation processing on the multi-source data to obtain contextual information related to the report topic from the multi-source data, and generating retrieval results based on the contextual information; and generating a target report based on the retrieval results. This solution, through retrieval enhancement generation technology, avoids information omissions and subjective interference compared to manual analysis, obtains accurate context, and ensures that the generated target report is based on accurate information, thereby improving the accuracy of the report. Attached Figure Description

[0026] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0027] Figure 1A schematic diagram illustrating an application scenario of a report generation method provided in this application embodiment;

[0028] Figure 2 A flowchart illustrating a report generation method provided in an embodiment of this application;

[0029] Figure 3 A flowchart illustrating another report generation method provided in this application embodiment;

[0030] Figure 4 A schematic diagram illustrating the generation of search results provided in an embodiment of this application;

[0031] Figure 5 A schematic diagram illustrating the generation of a combined report provided in an embodiment of this application;

[0032] Figure 6 A schematic diagram illustrating the generation of target analysis conclusions provided in an embodiment of this application;

[0033] Figure 7 A schematic diagram of a report generation apparatus provided in an embodiment of this application;

[0034] Figure 8 A schematic diagram of another report generation apparatus provided in an embodiment of this application;

[0035] Figure 9 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application.

[0036] The accompanying drawings illustrate specific embodiments of this application, which will be described in more detail below. These drawings and descriptions are not intended to limit the scope of the concept in any way, but rather to illustrate the concept of this application to those skilled in the art through reference to particular embodiments. Detailed Implementation

[0037] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.

[0038] In this application embodiment, "at least one" refers to one or more, and "more than one" refers to two or more. "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, or B alone, where A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. "At least one of the following" or similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one of a, b, or c can represent: a, b, c, ab, ac, bc, or abc, where a, b, and c can be single or multiple.

[0039] It should be noted that the phrase "at...time" in the embodiments of this application can refer to the instant at which a certain situation occurs, or to a period of time after the occurrence of a certain situation; the embodiments of this application do not specifically limit this. Furthermore, the display interface provided in the embodiments of this application is merely an example, and the display interface may include more or less content.

[0040] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties. Furthermore, the collection, use, processing, transmission, provision, disclosure, and application of related data all comply with relevant laws, regulations, and standards, necessary confidentiality measures have been taken, they do not violate public order and good morals, and corresponding operation entry points are provided for users to choose to authorize or refuse.

[0041] Furthermore, the technical solution involved in this application, which involves big data analysis of user information (including but not limited to personal biometrics, identity data, consumption data, asset data, electronic terminal operation data, etc.) and the use of artificial intelligence technology for automated decision-making, and makes decisions that have a significant impact on personal rights based on the results of automated decision-making, provides users with corresponding operation entry points for users to choose to agree to or reject the results of automated decision-making; if the user chooses to reject, the process will proceed to the expert decision-making process.

[0042] It should be noted that the method, apparatus, electronic device and storage medium for generating this report can be used in the field of artificial intelligence, or in any field other than artificial intelligence. The application fields of the method, apparatus, electronic device and storage medium for generating this report are not limited.

[0043] Figure 1This is a schematic diagram illustrating an application scenario of a report generation method provided in this application embodiment. An example is given based on the illustrated scenario: raw data is analyzed to obtain analysis results, and the analysis results are summarized and organized to obtain a report. The report can reflect the information corresponding to the raw data.

[0044] In related technologies, raw data is collected manually, analyzed, and finally a report is generated.

[0045] However, manual methods may miss important data and are subject to interference from the subjective factors of staff, resulting in low accuracy of the generated reports.

[0046] Using scenario examples, it's clear that in situations with diverse data sources and massive data volumes, manual data collection is prone to omissions. Manual analysis relies on the analyst's individual knowledge and experience; however, when faced with complex data, limitations in personal understanding or oversight can easily lead to overlooking crucial information or potential risks.

[0047] The method for generating reports provided in this application is intended to solve the above-mentioned technical problems of the prior art.

[0048] The technical solution of this application and how the technical solution of this application solves the above-mentioned technical problems are described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments. The embodiments of this application will now be described with reference to the accompanying drawings.

[0049] Figure 2 A flowchart illustrating a report generation method provided in this application embodiment, the method comprising the following steps:

[0050] S201, Received Report Subject.

[0051] The report title clearly defines the scope and core theme of the report to be generated.

[0052] For example, the report topic is the instruction issued by the user. The scope of the target report is accurately controlled by the report topic so that the generated target report meets the user's requirements.

[0053] S202. Obtain multi-source data, which includes structured and unstructured data from heterogeneous data sources.

[0054] For example, based on the received report topic, structured and unstructured data can be automatically collected from multiple independent sources with different data formats and structures.

[0055] For example, structured data is data with a predetermined structure and fixed fields, such as tables, JSON, and XML formatted data in a database. Unstructured data is text information without a predefined format, such as corporate annual reports, press releases, or industry analysis reports.

[0056] For example, the report generated by this application can be a financial or banking business report, and the multi-source data can be business-related data.

[0057] For example, automated data acquisition can be used to build a comprehensive, multi-source dataset, avoiding data omissions.

[0058] S203. Perform retrieval enhancement generation processing on multi-source data to obtain contextual information related to the report topic from the multi-source data, and generate retrieval results based on the contextual information.

[0059] For example, the retrieval enhancement generation technology combines the technical framework of retrieval systems and large language models, and solves the problems of knowledge solidification, high illusion rate and insufficient domain adaptability of large models through dynamic knowledge retrieval and generation collaboration.

[0060] For example, using the report topic as a query condition, semantic retrieval and other technologies are employed to proactively search and extract facts, data fragments, or text paragraphs most relevant to the report topic from multi-source data. The extracted information is referred to as contextual information.

[0061] For example, using contextual information as key input and factual constraints, analysis is performed, and search results are output. That is, the search results are the conclusions obtained from the preliminary analysis.

[0062] Based on the above implementation method, the context information is comprehensive and accurate, so accurate search results can be generated based on the context information.

[0063] S204. Generate a target report based on the search results.

[0064] For example, the search results are arranged and integrated according to the standardized format required by the business to obtain a complete and formal target report. The target report can accurately reflect the important information of the corresponding business.

[0065] For example, the process of generating a target report may also include operations such as inserting charts, generating summaries, adding page numbers and a table of contents.

[0066] The report generation method provided in this application includes receiving a report topic; acquiring multi-source data, including structured and unstructured data from heterogeneous data sources; performing retrieval enhancement generation processing on the multi-source data to obtain contextual information related to the report topic from the multi-source data, and generating retrieval results based on the contextual information; and generating a target report based on the retrieval results. This solution, through retrieval enhancement generation technology, avoids information omissions and subjective interference compared to manual analysis, obtains accurate context, and ensures that the generated target report is based on accurate information, thereby improving the accuracy of the report.

[0067] Based on any of the above embodiments, the following, in conjunction with Figure 3 The report provides a detailed explanation of the process by which it was generated.

[0068] Figure 3 This is a flowchart illustrating another report generation method provided in an embodiment of this application. Figure 3 As shown, the method includes:

[0069] S301, Receive report subject.

[0070] It should be noted that the execution process of S301 is the same as that of S201, and will not be repeated here.

[0071] S302. Obtain multi-source data, which includes structured and unstructured data from heterogeneous data sources.

[0072] It should be noted that the execution process of S302 is the same as that of S202, and will not be repeated here.

[0073] S303. Based on the report topic, perform semantic similarity retrieval in multi-source data to obtain a set of text fragments that are semantically related to the report topic from the multi-source data.

[0074] For example, using the report topic as the query criterion, semantic similarity calculation technology is employed to filter out the set of text fragments that are semantically closest from a large portion of unstructured data. Compared to simple keyword matching, this approach can understand the deeper meaning of complex concepts, thereby improving the accuracy of contextual information.

[0075] For example, for the structured data portion, detection can be performed directly through the data results and data fields of the structured data.

[0076] To illustrate with a scenario example, for instance, the report topic and multi-source data are both converted into high-dimensional vectors, and the cosine similarity of the high-dimensional vectors is calculated to perform semantic similarity retrieval.

[0077] Based on the above implementation methods, information directly relevant to the current analysis can be accurately located and extracted from messy unstructured data, thereby improving the accuracy of retrieval.

[0078] S304. The set of text fragments is used as an enhanced context and input into the pre-trained large language model to obtain the retrieval results. The large language model is trained based on historical texts.

[0079] For example, the filtered collection of text fragments is packaged as a whole and fed into a pre-trained large language model as enhanced context input. Based on its language understanding and generation capabilities gained from training on massive amounts of historical text, the large language model organizes language, performs analysis and reasoning, and creates text solely based on context, ultimately outputting structured analyzed text as the retrieval result.

[0080] Optional, large language model analysis modes include: risk cross-validation analysis mode, which extracts the enterprise's structured data and industry benchmark report data from the enhanced context, compares and cross-validates them, and generates risk analysis conclusions about the enterprise; compliance correlation analysis mode, which extracts the enterprise's event data and relevant regulations from the enhanced context, performs correlation analysis and compliance mapping, and generates compliance conclusions about the enterprise.

[0081] Below, in conjunction with Figure 4 Explanation of the generated search results.

[0082] Figure 4 This is a schematic diagram illustrating the generation of search results provided in an embodiment of this application. For example... Figure 4 As shown, multi-source data containing both structured and unstructured data is obtained. From the multi-source data, a set of text fragments semantically similar to the report's topic is retrieved. The retrieval results are obtained by inferring from the set of text fragments using a large language model.

[0083] Based on the above implementation method, by using the retrieved set of text fragments as enhanced context input to the large language model, the generation range of the large language model is strictly constrained, so that the large language model generates highly relevant search results, thereby improving the accuracy of the report.

[0084] S305. Obtain the initial report template, which includes objective data fields.

[0085] For example, the initial report template is a predefined structured framework.

[0086] For example, objective data fields are locations where precise numerical values, codes, or standardized text need to be entered.

[0087] One feasible implementation method is to obtain the initial report template by: receiving a service type identifier sent by the user terminal; obtaining the corresponding business rule from a preset business rule library based on the service type identifier; and dynamically generating the initial report template based on the business rule.

[0088] For example, when instructing the generation of a report, the user specifies a specific business type identifier. The business type identifier is a highly condensed, machine-readable business instruction that determines the analytical dimensions, data requirements, and risk concerns of the entire report.

[0089] For example, a business rule base is a structured knowledge base that stores requirements for different businesses, risk control logic, and analysis specifications. The business rule base transforms business knowledge into machine-executable rules.

[0090] For example, using the business type identifier as the key, the complete set of rules bound to it is retrieved from the rule base. These rules specifically define the data fields, analysis sections, risk assessment models, and specifications that the corresponding type of report must include.

[0091] For example, based on actual business rules, a structured document framework that meets all business rule requirements is built in real time, i.e., an initial report template. The initial report template embeds descriptions of chapter structure, field definitions, or some calculation logic.

[0092] In this feasible implementation, by encoding business rules into the generation logic of the initial report template, the initial report template can accurately match the business rules, avoiding misuse or omission that may occur when manually selecting templates, thereby improving the accuracy of the report.

[0093] S306. Using robotic process automation technology, extract standardized data corresponding to objective data fields from structured data.

[0094] Robotic Process Automation (RPA) is a technology that uses software robots and artificial intelligence to automatically perform repetitive office tasks.

[0095] For example, based on the definition of objective data fields, a series of predefined operations (such as logging into the system, querying the database, calling the application programming interface (API)) are automatically executed through robotic process automation technology to accurately extract the corresponding data from the structured data.

[0096] One feasible approach to extract standardized data involves the following steps: determining the data specification requirements corresponding to the objective data fields based on the initial report template; determining the data extraction rules and source mapping relationship based on the data specification requirements; and extracting the standardized data corresponding to the objective data fields from the structured data through an application programming interface based on the data extraction rules and source mapping relationship.

[0097] For example, parsing the initial report template not only identifies which objective data fields are in the initial report template, but also further clarifies the precise technical specifications of each field.

[0098] With the help of scenario examples, technical specifications can include: data type (e.g., numeric, percentage, date), data format (e.g., YYYY-MM-DD, two decimal places), unit of measurement, value range (e.g., 0-100), and possible calculation or conversion rules (e.g., taken from the consolidated profit statement, need to calculate the compound annual growth rate of the past three years).

[0099] Optionally, based on data specification requirements, a matching process can be performed in the configuration library to dynamically determine two key data items: data extraction rules and source mapping relationships.

[0100] For example, a data extraction rule is a specific operational instruction required to obtain data that meets the specified criteria. For instance, for the field "annual sales revenue," the data extraction rule might be: call the financial system interface to query the account "operating revenue," with the period being the most recent year.

[0101] For example, source mapping relationships are used to indicate the authoritative data source for each data field. This ensures, for instance, the authority and consistency of the data.

[0102] For example, based on the planned rules and mapping relationships, requests are initiated and responses are received in a programmatic and standardized manner by calling the standardized application programming interfaces provided by each data source to obtain non-standard data with inconsistent formats. Next, the non-standard data is strictly cleaned, transformed, and formatted according to preset requirements to output standardized data that meets the template input requirements.

[0103] In this feasible implementation, the precise technical specifications (such as numerical precision, date format, and unit of measurement) of each objective data field are parsed from the initial report template, and strict formatting is performed after extraction to avoid errors caused by manual methods due to data comprehension or inconsistent formats, thereby improving the accuracy of the report.

[0104] S307. Fill the standardized data into the positions corresponding to the objective data fields in the initial report template to obtain the intermediate report template.

[0105] For example, robotic process automation (RPA) technology automatically and accurately fills the captured data into the corresponding objective data fields according to the template requirements. This process replaces manual operation and can be executed accurately and quickly.

[0106] S308. Generate the target report based on the search results and intermediate report template.

[0107] For example, based on the intermediate report template, the search results are populated to ensure data integrity. Then, the data is processed according to the preset format and layout to obtain the correctly formatted target report.

[0108] One feasible implementation method is to generate the target report by: combining the search results with the intermediate report template to obtain a combined report; performing content consistency verification and data rationality verification on the combined report to obtain a verification result, which is either verification passed or verification failed; in response to the verification result being verification passed, formatting the combined report according to a preset report format to obtain the target report.

[0109] Below, in conjunction with Figure 5 Explanation of the generated portfolio report.

[0110] Figure 5 This is a schematic diagram illustrating the generation of a combined report provided in an embodiment of this application. Figure 5 As shown, robotic process automation (RPA) technology is used to extract standardized data corresponding to objective data fields from structured data. This standardized data represents the data required for the target report. The standardized data is then used to populate the initial report template to obtain an intermediate report template. Finally, the search results are used to populate the intermediate report template to obtain a combined report.

[0111] For example, the search results generated by the large language model are merged with the intermediate report template to generate a combined report that is complete but has not undergone final quality checks.

[0112] For example, content consistency checks focus on logical and semantic consistency. They examine whether arguments in the subjective analytical text of the combined report contradict objective data.

[0113] For example, data reasonableness verification focuses on the business reasonableness of the data. Based on pre-defined business rules and a risk control knowledge base, it assesses the reasonableness of key values ​​in the portfolio report. It can obtain data that does not conform to common business sense.

[0114] For example, a successful validation indicates that no major contradictions or anomalies were found in the combined report at the logical and data levels. The report is then automatically formatted according to a preset, standardized report format, outputting a directly usable target report.

[0115] Optionally, the report format may include, but is not limited to, at least one of the following: font, header and footer, table of contents, chart style, etc.

[0116] Optionally, if the verification fails, a report will not be output directly, but an alarm message can be output instead.

[0117] In this feasible implementation, automated verification can promptly detect and correct errors, thereby improving the accuracy of the report.

[0118] One feasible implementation method is to generate a combined report by: inputting the search results into multiple vertical analysis models to obtain multiple intermediate analysis conclusions, wherein the multiple vertical analysis models are obtained by training the models with sample data from different fields; fusing the multiple intermediate analysis conclusions through a preset decision fusion rule or model to obtain the target analysis conclusion; and filling the target analysis conclusion into the corresponding position of the intermediate report template to obtain the target report.

[0119] Below, in conjunction with Figure 6 Explain the conclusions of the target analysis.

[0120] Figure 6 This is a schematic diagram illustrating the generation of target analysis conclusions provided in an embodiment of this application. For example... Figure 6 As shown, the search results are analyzed using multiple vertical analysis models to obtain multiple intermediate analysis conclusions corresponding to multiple dimensions. These intermediate analysis conclusions are then fused to obtain the target analysis conclusion.

[0121] For example, the search results are not used directly. Instead, they are used as input and distributed to multiple vertical analysis models. Each vertical model is specifically trained using high-quality sample data from a particular domain. Each vertical model reviews and verifies the search results from its own professional perspective and outputs intermediate analysis conclusions.

[0122] Optionally, multiple vertical analysis models may include, but are not limited to, at least one of the following: financial analysis model, industry analysis model, etc.

[0123] For example, by fusing the intermediate analysis conclusions generated by multiple vertical analysis models, disagreements can be resolved, consensus can be reached, and the target analysis conclusion can be obtained.

[0124] Optionally, the fusion process includes cross-validation, conflict resolution, and weighted fusion operations.

[0125] In this feasible implementation, by integrating the intermediate analysis conclusions of multiple vertical analysis models, the bias or blind spots of a single model can be avoided, thereby improving the accuracy of the target report.

[0126] Figure 7This is a schematic diagram of a report generation apparatus provided in an embodiment of this application. Figure 7 As shown, the report generation device 70 may include: a receiving module 71, an acquisition module 72, a retrieval module 73, and a generation module 74.

[0127] The receiving module 71 is used to receive the report topic.

[0128] The acquisition module 72 is used to acquire multi-source data, which includes structured and unstructured data from heterogeneous data sources.

[0129] The retrieval module 73 is used to perform retrieval enhancement generation processing on multi-source data to obtain contextual information related to the report topic from the multi-source data and generate retrieval results based on the contextual information.

[0130] The generation module 74 is used to generate a target report based on the search results.

[0131] Optionally, the receiving module 71 can perform... Figure 2 S201 in the embodiment.

[0132] Optionally, module 72 can be executed. Figure 2 S202 in the embodiment.

[0133] Optionally, the retrieval module 73 can perform... Figure 2 S203 in the embodiment.

[0134] Optionally, the generation module 74 can be executed. Figure 2 S204 in the embodiment.

[0135] It should be noted that the report generation apparatus shown in the embodiments of this application can execute the technical solutions shown in the above method embodiments, and its implementation principle and beneficial effects are similar, so they will not be described again here.

[0136] In one possible implementation, the retrieval module 73 is specifically used for:

[0137] Based on the report topic, semantic similarity retrieval is performed in multi-source data to obtain a set of text fragments that are semantically related to the report topic from the multi-source data;

[0138] The set of text fragments is used as an enhanced context and input into a pre-trained large language model to obtain retrieval results. The large language model is trained based on historical texts.

[0139] In one possible implementation, the generation module 74 is specifically used for:

[0140] Obtain the initial report template, which includes objective data fields;

[0141] Robotic process automation (RPA) technology is used to extract standardized data corresponding to objective data fields from structured data.

[0142] The standardized data is then filled into the corresponding positions of the objective data fields in the initial report template to obtain the intermediate report template.

[0143] Generate the target report based on the search results and intermediate report template.

[0144] In one possible implementation, the generation module 74 is specifically used for:

[0145] Receive the service type identifier sent by the user terminal;

[0146] Based on the business type identifier, retrieve the corresponding business rules from the preset business rule library;

[0147] Based on business rules, an initial report template is dynamically generated.

[0148] Figure 8 This is a schematic diagram of another report generation apparatus provided in an embodiment of this application. Figure 7 Based on the illustrated embodiments, as Figure 8 As shown, the report generation device 70 also includes an extraction module 75 and an execution module 76.

[0149] Extraction module 75, used for:

[0150] Based on the initial report template, determine the data specification requirements corresponding to the objective data fields;

[0151] Based on the data specifications, determine the data extraction rules and source mapping relationships;

[0152] Based on the data extraction rules and source mapping relationship, standardized data corresponding to objective data fields are extracted from structured data through application programming interfaces.

[0153] Execution module 76 is used for:

[0154] Combine the search results with the intermediate report template to obtain a combined report;

[0155] Perform content consistency verification and data rationality verification on the combined report, and obtain the verification results, which are either verification passed or verification failed.

[0156] In response to a successful verification result, the combined report is formatted according to a preset report format to obtain the target report.

[0157] In one possible implementation, execution module 76 is specifically used for:

[0158] The search results were input into multiple vertical analysis models to obtain multiple intermediate analysis conclusions. These multiple vertical analysis models were obtained by training the models with sample data from different fields.

[0159] By using pre-defined decision fusion rules or models, multiple intermediate analysis conclusions are fused to obtain the target analysis conclusion;

[0160] Fill the target analysis conclusions into the corresponding positions in the intermediate report template to obtain the target report.

[0161] Figure 9 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application, such as... Figure 9 As shown, the electronic device includes:

[0162] The electronic device includes a processor 291 and a memory 292; it may also include a communication interface 293 and a bus 294. The processor 291, memory 292, and communication interface 293 can communicate with each other via the bus 294. The communication interface 293 can be used for information transmission. The processor 291 can invoke logical instructions stored in the memory 292 to execute the methods of the above embodiments.

[0163] Furthermore, the logic instructions in the aforementioned memory 292 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium.

[0164] The memory 292, as a non-volatile computer-readable storage medium, can be used to store software programs and computer-executable programs, such as program instructions / modules corresponding to the methods in the embodiments of this application. The processor 291 executes functional applications and data processing by running the software programs, instructions, and modules stored in the memory 292, that is, it implements the methods in the above-described method embodiments.

[0165] The memory 292 may include a program storage area and a data storage area. The program storage area may store the operating system and application programs required for at least one function; the data storage area may store data created based on the use of the terminal device. Furthermore, the memory 292 may include high-speed random access memory and may also include non-volatile memory.

[0166] This application provides a non-volatile computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, are used to implement the method as described in the foregoing embodiments.

[0167] This application provides a computer program product, including a computer program that, when executed by a processor, implements the method as described in the foregoing embodiments.

[0168] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, as some steps may be performed in other orders or simultaneously according to this application. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are all optional embodiments, and the actions and modules involved are not necessarily essential to this application.

[0169] It should be further noted that although the steps in the flowchart are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps; they can be executed in other orders. Moreover, at least some steps in the flowchart may include multiple sub-steps or multiple stages, which do not necessarily complete at the same time but can be executed at different times. The execution order of these sub-steps or stages is also not necessarily sequential but can be alternated or carried out in turn with other steps or at least some of the sub-steps or stages of other steps.

[0170] It should be understood that the above-described device embodiments are merely illustrative, and the device of this application can also be implemented in other ways. For example, the division of units / modules in the above embodiments is only a logical functional division, and there may be other division methods in actual implementation. For example, multiple units, modules, or components may be combined, or integrated into another system, or some features may be ignored or not executed.

[0171] Furthermore, unless otherwise specified, the functional units / modules in the various embodiments of this application can be integrated into one unit / module, or each unit / module can exist physically separately, or two or more units / modules can be integrated together. The integrated units / modules described above can be implemented in hardware or as software program modules.

[0172] When the integrated unit / module is implemented in hardware, the hardware can be digital circuits, analog circuits, etc. The physical implementation of the hardware structure includes, but is not limited to, transistors, memristors, etc. The processor can be any suitable hardware processor, such as CPU, GPU, FPGA, DSP, and ASIC. The storage unit can be any suitable magnetic or magneto-optical storage medium, such as Resistive Random Access Memory (RRAM), Dynamic Random Access Memory (DRAM), Static Random Access Memory (SRAM), Enhanced Dynamic Random Access Memory (EDRAM), High-Bandwidth Memory (HBM), Hybrid Memory Cube (HMC), etc.

[0173] If the integrated unit / module is implemented as a software program module and sold or used as an independent product, it can be stored in a computer-readable storage device (CMD). Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a memory and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned memory includes various media capable of storing program code, such as a USB flash drive, read-only memory (ROM), random access memory (RAM), portable hard drive, magnetic disk, or optical disk.

[0174] In the above embodiments, the descriptions of each embodiment have their own emphasis. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments. The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as the combination of these technical features does not contradict each other, it should be considered within the scope of this specification.

[0175] Other embodiments of this application will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of this application that follow the general principles of this application and include common knowledge or customary techniques in the art not disclosed herein. The specification and examples are to be considered exemplary only, and the true scope and spirit of this application are indicated by the claims.

[0176] It should be understood that this application is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this application is limited only by the appended claims.

Claims

1. A method for generating a report, characterized in that, include: Received report subject; Acquire multi-source data, which includes structured and unstructured data from heterogeneous data sources; The multi-source data is subjected to retrieval enhancement and generation processing to obtain contextual information related to the report topic from the multi-source data, and retrieval results are generated based on the contextual information. Based on the search results, a target report is generated.

2. The method according to claim 1, characterized in that, Perform retrieval enhancement generation processing on the multi-source data to obtain contextual information related to the report topic from the multi-source data, and generate retrieval results based on the contextual information, including: Based on the report topic, semantic similarity retrieval is performed in the multi-source data to obtain a set of text fragments that are semantically related to the report topic from the multi-source data; The set of text fragments is used as an enhanced context and input into a pre-trained large language model to obtain the retrieval results. The large language model is trained based on historical text.

3. The method according to claim 1, characterized in that, Based on the search results, a target report is generated, including: Obtain an initial report template, which includes objective data fields; Robotic process automation (RPA) technology is used to extract standardized data corresponding to the objective data fields from the structured data. The standardized data is filled into the positions corresponding to the objective data fields in the initial report template to obtain the intermediate report template; The target report is generated based on the search results and the intermediate report template.

4. The method according to claim 3, characterized in that, Obtain the initial report template, including: Receive the service type identifier sent by the user terminal; Based on the business type identifier, retrieve the corresponding business rules from the preset business rule base; The initial report template is dynamically generated based on the business rules.

5. The method according to claim 3, characterized in that, Extracting standardized data corresponding to the objective data fields from the structured data specifically includes: Based on the initial report template, determine the data specification requirements corresponding to the objective data fields; Based on the data specifications, determine the data extraction rules and source mapping relationship; Based on the data extraction rules and source mapping relationship, standardized data corresponding to the objective data fields are extracted from the structured data through the application programming interface.

6. The method according to claim 3, characterized in that, Based on the search results and the intermediate report template, the target report is generated, including: The search results are combined with the intermediate report template to obtain a combined report; The combined report is subjected to content consistency verification and data rationality verification to obtain verification results, which are either verification passed or verification failed. In response to the verification result being successful, the combined report is formatted according to a preset report format to obtain the target report.

7. The method according to claim 6, characterized in that, The search results are combined with the intermediate report template to obtain a combined report, including: The search results are input into multiple vertical analysis models to obtain multiple intermediate analysis conclusions. The multiple vertical analysis models are obtained by training the models with sample data from different fields. By using preset decision fusion rules or models, the multiple intermediate analysis conclusions are fused to obtain the target analysis conclusion; The target analysis conclusions are filled into the corresponding positions in the intermediate report template to obtain the target report.

8. A report generation apparatus, characterized in that, include: The receiving module is used to receive report topics; The acquisition module is used to acquire multi-source data, which includes structured data and unstructured data from heterogeneous data sources; The retrieval module is used to perform retrieval enhancement generation processing on the multi-source data to obtain contextual information related to the report topic from the multi-source data, and generate retrieval results based on the contextual information; The generation module is used to generate a target report based on the search results.

9. An electronic device, characterized in that, include: A processor, and a memory communicatively connected to the processor; The memory stores computer-executed instructions; The processor executes computer execution instructions stored in the memory to implement the method as described in any one of claims 1-7.

10. A non-volatile computer-readable storage medium, characterized in that, The non-volatile computer-readable storage medium stores computer-executable instructions, which, when executed by a processor, are used to implement the method as described in any one of claims 1-7.

Images