Document generation method, document generation device, electronic equipment and readable medium

Abstract

The invention provides a document generation method and device, electronic equipment and a computer readable medium, and belongs to the technical field of computers. The method comprises the steps of determining an associated subject corresponding to a to-be-generated target document, and generating a data processing task according to associated data corresponding to the associated subject; distributing the data processing task to a preset first intelligent agent, so that the first intelligent agent processes the associated data according to the data processing task to generate document reference data; inputting the document reference data into a preset second intelligent agent, so that the second intelligent agent evaluates the document reference data to obtain an evaluation result of the document reference data; and generating a target document according to an evaluation result of the document reference data. According to the embodiment of the invention, the automatic document generation can be realized through the cooperation of the intelligent agents with different division of labor.

Description

Technical Field

[0001] The present application relates to the field of computer technology, and in particular to a document generation method, a document generation device, an electronic device, and a computer-readable storage medium. Background Art

[0002] Research document writing is involved in all fields. For example, in the financial field, it is necessary to write corresponding research reports based on the application scenario requirements. Research reports refer to research reports that analyze financial market conditions, industries, macro strategies, etc.

[0003] At present, research documents in various fields are mainly written based on manual experience. For example, research reports in the financial field are mainly written based on researchers' experience and understanding of market conditions. However, the above manual writing method is difficult to generate documents on a large scale and automatically, resulting in low document generation efficiency. Summary of the invention

[0004] The present application provides a document generation method, a document generation device, an electronic device, and a computer-readable storage medium, which can realize automated document generation through the division of labor and cooperation of multiple agents.

[0005] In a first aspect, the present application provides a document generation method, the document generation method comprising:

[0006] Determine the associated subject corresponding to the target document to be generated, and generate a data processing task according to the associated data corresponding to the associated subject;

[0007] Allocating the data processing task to a preset first agent, so that the first agent processes the associated data according to the data processing task to generate document reference data;

[0008] Inputting the document reference data into a preset second agent, so that the second agent evaluates the document reference data to obtain an evaluation result of the document reference data;

[0009] The target document is generated according to the evaluation result of the document reference data.

[0010] In a second aspect, the present application provides a document generation device, the document generation device comprising:

[0011] A task construction module, used to determine the associated subject corresponding to the target document to be generated, and generate a data processing task according to the associated data corresponding to the associated subject;

[0012] A data generation module, used for allocating the data processing task to a preset first agent, so that the first agent processes the associated data according to the data processing task to generate document reference data;

[0013] A data evaluation module, used for inputting the document reference data into a preset second agent, so that the second agent evaluates the document reference data to obtain an evaluation result of the document reference data;

[0014] The document generation module is used to generate the target document according to the evaluation result of the document reference data.

[0015] In a third aspect, the present application provides an electronic device comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores one or more computer programs executable by the at least one processor, and one or more of the computer programs are executed by the at least one processor so that the at least one processor can execute the above-mentioned document generation method.

[0016] In a fourth aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program implements the above-mentioned document generation method when executed by a processor / processing core.

[0017] In the embodiment provided by the present application, for the target document to be generated, a data processing task is generated through the associated data corresponding to its associated subject, so that the preset first intelligent agent is responsible for executing the data processing task, processing the associated data, and generating document reference data. Then, the preset second intelligent agent further evaluates the document reference data generated by the first intelligent agent to obtain the evaluation result of the document reference data, and finally generates the target document according to the evaluation result of the document reference data. As can be seen from the above, on the one hand, the embodiment of the present application realizes automatic document generation by setting preset intelligent agents with different divisions of labor to cooperate, thereby improving the efficiency of document generation. On the other hand, the embodiment of the present application further evaluates the document reference data based on the second intelligent agent, thereby generating the target document according to the evaluation result, so that the content of the generated target document is more accurate and reliable.

[0018] It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present application, nor is it intended to limit the scope of the present application. Other features of the present application will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The accompanying drawings are used to provide a further understanding of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. The above and other features and advantages will become more apparent to those skilled in the art by describing detailed example embodiments with reference to the accompanying drawings. In the accompanying drawings:

[0020] Figure 1 A flowchart of a document generation method provided in an embodiment of the present application;

[0021] Figure 2 A flowchart of a document generation method provided in an embodiment of the present application;

[0022] Figure 3 A schematic diagram of a document generation method provided in an embodiment of the present application;

[0023] Figure 4 A block diagram of a document generation device provided in an embodiment of the present application;

[0024] Figure 5 A block diagram of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0025] In order to enable those skilled in the art to better understand the technical solution of the present application, the exemplary embodiments of the present application are described below in conjunction with the accompanying drawings, including various details of the embodiments of the present application to facilitate understanding, which should be considered as merely exemplary. Therefore, those of ordinary skill in the art should recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of the present application. Similarly, for the sake of clarity and conciseness, the description of well-known functions and structures is omitted in the following description.

[0026] In the absence of conflict, the embodiments of the present application and the features therein may be combined with each other.

[0027] As used herein, the term "and / or" includes any and all combinations of one or more of the associated listed items.

[0028] The terms used herein are only used to describe specific embodiments and are not intended to limit the present application. As used herein, the singular forms "a" and "the" are also intended to include plural forms, unless the context clearly indicates otherwise. It will also be understood that when the terms "including" and / or "made of" are used in this specification, the presence of the features, wholes, steps, operations, elements and / or components is specified, but the presence or addition of one or more other features, wholes, steps, operations, elements, components and / or groups thereof is not excluded. "Connected" or "connected" and similar words are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect.

[0029] Unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by those of ordinary skill in the art. It will also be understood that terms such as those defined in commonly used dictionaries should be interpreted as having a meaning consistent with their meaning in the context of the relevant art and this application, and will not be interpreted as having an idealized or overly formal meaning unless explicitly defined herein.

[0030] In related technologies, research documents in different fields are mainly written by professionals in the field based on their analysis and understanding, resulting in low document generation efficiency. For example, research reports in the financial field are mainly written by researchers in the financial field based on their work experience and understanding of market conditions.

[0031] In view of this, the embodiments of the present application provide a document generation method, a document generation device, an electronic device, and a readable medium. For a target document to be generated, a data processing task is generated through the associated data corresponding to its associated subject, so that a preset first intelligent agent is responsible for executing the data processing task, processing the associated data, and generating document reference data. Then, a preset second intelligent agent further evaluates the document reference data generated by the first intelligent agent to obtain an evaluation result of the document reference data, and finally generates the target document based on the evaluation result of the document reference data.

[0032] From the above, it can be concluded that, on the one hand, the embodiment of the present application realizes automated document generation by setting up preset intelligent agents with different divisions of labor to collaborate, thereby improving the efficiency of document generation. On the other hand, the embodiment of the present application further evaluates the document reference data based on the second intelligent agent, and generates a target document based on the evaluation results, so that the content of the generated target document is more accurate and reliable.

[0033] The document generation method according to the embodiment of the present application can be performed by an electronic device such as a terminal device or a server, and the terminal device can be a user equipment (User Equipment, UE), a mobile device, a user terminal, a terminal, a cellular phone, a cordless phone, a personal digital assistant (Personal Digital Assistant, PDA), a handheld device, a computing device, a vehicle-mounted device, a wearable device, etc. The method can be implemented by a processor calling a computer-readable program instruction stored in a memory. The server can be an independent physical server, a server cluster composed of multiple servers, or a cloud server capable of cloud computing.

[0034] Figure 1 A flowchart of a document generation method provided in an embodiment of the present application. Figure 1 , the method comprising:

[0035] Step S110: Determine the associated subject corresponding to the target document to be generated, and generate a data processing task according to the associated data corresponding to the associated subject.

[0036] The target document to be generated may be a research document corresponding to various fields, such as a research report in the financial field, an analysis report corresponding to multiple fields such as the real estate field and the Internet field, which is not limited in the present embodiment of the application.

[0037] The associated subject corresponding to the target document to be generated refers to the research object corresponding to the target document to be generated, for example, a company, enterprise, product, project, market, etc., which is not limited in the present embodiment of the application, and thus, the target document to be generated is a report obtained by analyzing the research object.

[0038] In an optional implementation, the document generation request may be parsed to determine the associated subject corresponding to the target document to be generated.

[0039] The document generation request may be in various forms such as directive and question-based. For example, a question-based document generation request may be a document writing question raised by a user, such as "How is the market performance of a certain stock?", while a directive document generation request may be a document writing command directly input by a user, such as "Please generate a corresponding research report for a certain stock." Thus, by parsing the above document generation request, it can be obtained that the associated subject corresponding to the target document to be generated is a certain stock.

[0040] It should also be noted that when parsing the document generation request, the associated intent corresponding to the associated subject can also be determined. The associated intent is used to indicate the specific research content of the associated subject corresponding to the target document to be generated. For example, in the above document generation request, its associated intent can be stock market analysis.

[0041] Correspondingly, the associated data corresponding to the associated subject is attribute data related to the associated subject. According to the associated data, the associated subject can be effectively analyzed.

[0042] In an optional implementation, in order to improve the generation effect of the target document, the associated subject may correspond to associated data of multiple data types.

[0043] The data type is used to identify the data attribute corresponding to the associated data. Specifically, the data type may include a text data type and a time series data type.

[0044] The associated data corresponding to the text data type is the text data related to the associated subject. Further, the text data type can be divided into news data type, document data type, basic facet data type, etc. according to different data attributes, which is not limited in the present embodiment.

[0045] Among them, when the data type is a news data type, the associated data includes news information related to the associated subject, for example, market news information of a certain stock, news information of a company corresponding to a certain stock, etc.

[0046] When the data type is a document data type, the associated data includes historical document data related to the associated subject, such as historical research reports related to a certain stock, historical analysis reports on a certain product, and research reports related to a certain market.

[0047] When the data type is a fundamental data type, the associated data includes the fundamental data related to the associated subject. Fundamental data is data used to characterize the basic situation of the associated subject, such as a company, industry, etc. For example, company financial data, operating data, industry data, etc.

[0048] Among them, the associated data corresponding to the time series data type is the time series data related to the associated subject. Time series data contains a series of data arranged in chronological order. Technical analysis of time series data through relevant technical indicators can reflect the trend changes corresponding to the associated subject.

[0049] Furthermore, the time series data type can be divided into technical data types such as market data type, sales data type, production data type, etc. according to different data attributes.

[0050] It should be noted that under different associated subjects, the associated data of the corresponding time series data type may be different. For example, when the associated subject is a certain stock, the associated data of the time series data type may be stock data. When the associated subject is a certain product, the associated data of the time series data type may be the product sales data or product production data of the product, etc., which is not limited in the embodiments of the present application.

[0051] Since the associated data corresponds to different data types, different associated data has its corresponding data characteristics. Therefore, according to the associated data corresponding to the associated subject, it is necessary to generate the corresponding data processing task for the associated data, so as to perform targeted data processing on the associated data.

[0052] The data processing task is used to indicate the corresponding data processing operation to be performed on the associated data. For example, for the associated data of a certain news data type, the corresponding data processing operation may be data aggregation processing.

[0053] In an optional implementation, in order to improve the generation efficiency of the data processing task, the associated data and the corresponding data processing operation can be filled into the task instruction template to obtain the data processing task.

[0054] Exemplarily, the task instruction template may include associated data to be filled in, data processing operations to be filled in, and preset task content, such as "Please perform data processing on the following associated data according to the following data processing operations."

[0055] It should be noted that, since the associated data may correspond to multiple data types, the associated data of different data types may correspond to data processing operations of multiple operation types. The operation type is specifically used to characterize the data processing form corresponding to the data processing operation.

[0056] Step S120: Allocate the data processing task to a preset first agent, so that the first agent processes the associated data according to the data processing task to generate document reference data.

[0057] The preset first agent refers to an agent built based on a large language model. Specifically, the preset first agent can be an existing large language model, or can be obtained by fine-tuning the existing large language model, which is not limited in the present embodiment of the application.

[0058] It should be noted that the number of the first agents can be one or more, and this embodiment of the present application does not limit this.

[0059] Thus, by allocating the data processing task to the preset first agent, the first agent can process the associated data according to the data processing task to generate document reference data.

[0060] The document reference data is the processing result obtained after the first agent processes the associated data.

[0061] Specifically, when the associated data includes associated data corresponding to multiple data types, document reference data corresponding to multiple reference data types will be generated, wherein the reference data type is determined according to the data type of the associated data.

[0062] For example, if the associated data is associated data corresponding to a news data type, then the reference data type of the document reference data generated by the first agent is also a news data type.

[0063] Step S130: input the document reference data into a preset second agent, so that the second agent evaluates the document reference data to obtain an evaluation result of the document reference data.

[0064] The preset second agent refers to an agent built based on a large language model. Specifically, the preset second agent can be an existing large language model, or can be obtained by fine-tuning the existing large language model, which is not limited in the present embodiment of the application.

[0065] Specifically, the preset second intelligent agent is used to evaluate the document reference data, for example, to evaluate and analyze the document reference data according to the association intention corresponding to the association subject, so as to generate conclusive evaluation content corresponding to the association intention.

[0066] Exemplarily, if the associated intent corresponding to the associated subject is stock market analysis, the second agent is used to evaluate the input document reference data such as stock analysis data, thereby obtaining evaluation content corresponding to the associated intent, for example, stock market analysis content.

[0067] Correspondingly, the evaluation result of the document reference data is the evaluation result corresponding to the associated intention obtained by the second intelligent agent evaluating the document reference data according to the associated intention corresponding to the associated subject.

[0068] It should be noted that the number of the second agents can be one or more, and this embodiment of the present application does not limit this.

[0069] Step S140: Generate a target document according to the evaluation result of the document reference data.

[0070] The target document can be generated by filling the evaluation results of the document reference data into the corresponding document template. The evaluation results of the document reference data can also be directly input into a preset agent for generating the target document, so that the agent directly generates the target document, which is not limited in the present embodiment.

[0071] The document generation method provided in the embodiment of the present application realizes automatic document generation by setting preset intelligent agents with different division of labor to collaborate, thereby improving the efficiency of document generation. On the other hand, the embodiment of the present application further evaluates the document reference data based on the second intelligent agent, and generates a target document according to the evaluation results, so that the content of the generated target document is more accurate and reliable.

[0072] In addition, those skilled in the art may also make various modifications and variations to the method in this application:

[0073] In an optional implementation, in order to improve the generation quality of the evaluation results of the document reference data, multiple second agents corresponding to different preset roles can be set according to different preset roles, and different preset roles correspond to different evaluation dimensions. Different second agents are set in the same group chat environment, so that the second agents corresponding to different preset roles can evaluate the document reference data according to the evaluation dimensions corresponding to the preset roles to obtain the evaluation results of the document reference data.

[0074] Correspondingly, the preset second agent includes: multiple second agents corresponding to different preset roles, and different preset roles correspond to different evaluation dimensions; and multiple second agents are set in the same group chat environment, and the document reference data is input into the preset second agent so that the second agent evaluates the document reference data to obtain the evaluation result of the document reference data, including: inputting the document reference data into multiple second agents so that the multiple second agents evaluate the document reference data in multiple rounds of interaction in the group chat environment according to the evaluation dimensions corresponding to the second agents, and obtain multiple rounds of interaction data between the multiple second agents; based on the multiple rounds of interaction data, obtain the evaluation result corresponding to the document reference data.

[0075] The preset role is used to represent the responsibilities assigned to the second agent. Specifically, the preset role can be determined according to the document type of the target document to be generated. For example, if the document type of the target document to be generated is a financial research report type, the preset role can be used for a financial analyst.

[0076] It should be noted that the evaluation dimension is used to characterize the evaluation method of the second agent on the document reference data.

[0077] In an optional implementation, the evaluation dimensions may include evaluation tendency, evaluation data type and / or evaluation level.

[0078] The evaluation tendency is used to characterize the emotional tendency of the second agent when evaluating the document reference data. For example, the evaluation tendency can be an optimistic evaluation tendency, a pessimistic evaluation tendency, or a neutral evaluation tendency.

[0079] The evaluation data type is used to characterize the reference data type corresponding to the document reference data evaluated by the second agent. For example, a second agent A is only used to evaluate document reference data corresponding to the news data type and document reference data corresponding to the time series data type, while a second agent B is used to evaluate document reference data corresponding to all reference data types, and this embodiment of the application does not limit this.

[0080] The evaluation level is used to represent the evaluation angle corresponding to the second agent when evaluating the document reference data. For example, the evaluation level can be a macro-type evaluation level for evaluating from a macro perspective, a micro-type evaluation level for evaluating from a micro perspective, an indicator-type evaluation level for evaluating from an evaluation indicator perspective, etc.

[0081] Therefore, by setting different evaluation dimensions for different preset roles, the second agent can evaluate the document reference data from different evaluation dimensions, thereby obtaining a more comprehensive evaluation result of the document reference data.

[0082] The group chat environment refers to an interactive environment corresponding to multiple second intelligent agents, and multiple second intelligent agents interact in the group chat environment to obtain interactive data between the multiple second intelligent agents.

[0083] Specifically, the interaction termination condition may be set so that when the interaction termination condition is reached between the plurality of second agents, the interaction is determined to be terminated. For example, when the interaction rounds between the plurality of second agents reach a preset interaction round, the interaction is determined to be terminated.

[0084] In each interaction round, each second agent evaluates the document reference data according to its corresponding evaluation dimension in the group chat environment. In addition, each second agent can also discuss the speeches of other second agents, and finally obtain the interaction data corresponding to the interaction round.

[0085] Therefore, after multiple rounds of interaction are completed, the evaluation results corresponding to the document reference data can be obtained based on the multiple rounds of interaction data.

[0086] In an embodiment of the present application, by setting up second intelligent agents corresponding to different preset roles, the second intelligent agents can evaluate the document reference data from different evaluation dimensions, so that based on multiple rounds of interactions between the second intelligent agents, evaluation results corresponding to the document reference data are obtained, so that the evaluation results can integrate diverse perspectives and richer content.

[0087] Furthermore, since the second agent can specifically evaluate the document reference data corresponding to one or more reference data types, in order to achieve accurate correspondence between the document reference data and the second agent, the second agent used to evaluate the document reference data can be determined based on the evaluation dimension corresponding to the second agent.

[0088] In an optional implementation, the document reference data includes multiple document reference data corresponding to multiple reference data types, and each second agent is used to evaluate the document reference data corresponding to at least one reference data type according to the corresponding evaluation dimension. Then, inputting the document reference data into a preset second agent includes: determining the target reference data type of the document reference data to be evaluated by each second agent according to the evaluation dimension corresponding to each second agent; for each document reference data, matching the reference data type of the document reference data with the target reference data type corresponding to each second agent, and determining the target agent corresponding to the document reference data from multiple second agents according to the matching result; for each document reference data, inputting the document reference data into the target agent corresponding to the document reference data.

[0089] The target reference data type is the reference data type of the document reference data that can be evaluated by the second agent. The target reference data type can be determined according to the evaluation dimension corresponding to the second agent.

[0090] Specifically, when the evaluation dimension corresponding to the second agent is an evaluation data type, the evaluation data type represents the reference data type of the document reference data that can be evaluated by the second agent.

[0091] Thus, for each document reference data, by matching its reference data type with the target reference data type corresponding to each second agent, when the matching result is a successful match, the second agent is determined as the target agent for evaluating the document reference data.

[0092] Therefore, by inputting the document reference data into its corresponding target intelligent agent, the target intelligent agent can evaluate the document reference data in a targeted manner, thereby improving the generation quality of the evaluation result.

[0093] In an optional implementation, in order to further improve the content accuracy of the evaluation results corresponding to the document reference data, a third intelligent agent can be set in the group chat environment to extract key interaction data from each round of interaction data based on the third intelligent agent, and obtain the evaluation results corresponding to the document reference data based on the key interaction data.

[0094] Correspondingly, a third agent is set up in the group chat environment to obtain evaluation results corresponding to the document reference data based on multiple rounds of interaction data, including: extracting data for each round of interaction data based on the third agent to obtain key interaction data corresponding to multiple second agents in each interaction round; for each interaction round, generating evaluation results corresponding to each interaction round based on the key interaction data corresponding to multiple second agents; and obtaining evaluation results corresponding to the document reference data based on the evaluation results corresponding to each interaction round.

[0095] The third agent and the second agent are set in the same group chat environment to perform data analysis on the interaction data between the second agents.

[0096] The third agent refers to an agent built based on a large language model. Specifically, the third agent can be an existing large language model, or can be obtained by fine-tuning the existing large language model, which is not limited in the present embodiment of the application.

[0097] Among them, the key interaction data refers to the opinion and / or summary interaction data output by multiple second agents in each interaction round.

[0098] Specifically, by inputting prompt information to the third agent, such as "please monitor the speech of each agent in each interaction round and extract opinion speech information", the third agent can extract key interaction data from the interaction data in each interaction round according to the prompt information.

[0099] In an optional implementation, the document reference data includes multiple document reference data corresponding to multiple reference data types, and each second agent is used to evaluate the document reference data corresponding to at least one reference data type according to the corresponding evaluation dimension, then the data extraction based on the third agent for each round of interaction data to obtain the key interaction data corresponding to the multiple second agents in each interaction round includes:

[0100] Determine, according to the evaluation dimension corresponding to each second agent, a second agent for evaluating the document reference data corresponding to each reference data type;

[0101] For each interaction round, based on the third agent, the interaction data corresponding to the second agent for evaluating each document reference data is respectively extracted from the interaction data of each round, and the key interaction data corresponding to each document reference data in each interaction round is determined according to the interaction data, so as to obtain the key interaction data corresponding to the multiple document reference data in each interaction round;

[0102] For each interaction round, the key interaction data corresponding to each interaction round is obtained according to the key interaction data corresponding to the multiple document reference data in each interaction round.

[0103] Specifically, when the evaluation dimension is the evaluation data type, the second agent used to evaluate the document reference data corresponding to each reference data type can be directly determined according to the evaluation data type. For example, the evaluation data types of the second agent A are reference data type 1 and reference data type 2, and the evaluation data types of the second agent B are reference data type 2 and reference data type 3. Therefore, the second agent used to evaluate the document reference data corresponding to reference data type 1 is the second agent A.

[0104] Therefore, when the third agent generates the key interaction data corresponding to each interaction round, it needs to extract the interaction data corresponding to the second agent in each round of interaction data according to the second agent corresponding to each document reference data, and then determine the key interaction data corresponding to the document reference data according to the interaction data. Finally, the key interaction data corresponding to each interaction round can be obtained according to the key interaction data corresponding to multiple document reference data.

[0105] In an embodiment of the present application, the interaction data of the second intelligent agent used to evaluate each type of document reference data is extracted from the interaction data by a third intelligent agent, so that key interaction data corresponding to the document reference data is obtained based on the interaction data, so that the key interaction data corresponding to each interaction round finally generated can cover more comprehensive and accurate content.

[0106] It should be noted that, in each interaction round, each second agent has its corresponding key interaction data, so according to the key interaction data of each second agent in the interaction round, the evaluation result corresponding to each interaction round can be generated.

[0107] Finally, the evaluation results corresponding to the document reference data can be obtained according to the evaluation results corresponding to each interaction round. For example, the evaluation results corresponding to each interaction round are combined to obtain the evaluation results corresponding to the document reference data, or the evaluation results corresponding to each interaction round are summarized to obtain a summary evaluation result corresponding to the document reference data.

[0108] In an embodiment of the present application, by setting up a third agent in a group chat environment, the third agent can further extract key interaction data from each round of interaction data, thereby generating evaluation results corresponding to the document reference data based on the key interaction data, thereby improving the content accuracy of the evaluation results.

[0109] In an optional implementation, in order to improve the generation effect of the target document, a document template corresponding to the target document can be generated according to the reference data type corresponding to the document reference data, thereby generating a target research report based on the document template and the evaluation results of the document reference data.

[0110] Correspondingly, the evaluation results of the document reference data include evaluation results of multiple document reference data corresponding to multiple reference data types in multiple interaction rounds. Then, a target document is generated based on the evaluation results of the document reference data, including: determining multiple document sections corresponding to the multiple reference data types according to the multiple reference data types corresponding to the multiple document reference data; generating a document template corresponding to the target document to be generated based on the multiple document sections; for the evaluation results of the multiple document reference data corresponding to the multiple reference data types in each interaction round, filling the evaluation results of the document reference data corresponding to each reference data type into the document section corresponding to the reference data type in the document template, to generate an initial document corresponding to each interaction round; generating a target document based on the multiple initial documents corresponding to the multiple interaction rounds.

[0111] The document section is a section structure included in the document template. Specifically, the document section may include a section title and a section area to be filled.

[0112] Among them, according to each reference data type, the document section corresponding to each reference data type can be determined. Specifically, according to each reference data type, the section title corresponding to the document section can be determined. For example, when the reference data type is a news data type, the corresponding section title can be news data analysis. And according to the section title, the corresponding section area is determined, for example, in the upper left area of ​​the document, which is not limited in the embodiment of the present application.

[0113] Thus, the document section corresponding to the reference data type can be determined according to the section title and the section area to be filled.

[0114] Furthermore, based on the multiple document sections, a document template corresponding to the target document to be generated can be generated.

[0115] Thus, for each interaction round, the initial document corresponding to each interaction round can be generated by filling the evaluation results of the document reference data corresponding to each reference data type into the document section corresponding to the reference data type in the document template.

[0116] Specifically, when filling the document section, it is necessary to fill the evaluation results of the document reference data into the section area to be filled in the document section corresponding to the reference data type.

[0117] Since the evaluation opinions output by multiple second agents may be continuously optimized during the continuous interaction process, the target document is generated based on multiple initial documents corresponding to multiple interaction rounds, so the content of the generated target document is more accurate and reliable.

[0118] In an optional implementation, in order to improve the document quality of the generated target document, the target document may be generated by a preset fourth agent based on multiple initial documents corresponding to multiple interaction rounds.

[0119] Correspondingly, generating the target document based on multiple initial documents corresponding to multiple interaction rounds includes: inputting the multiple initial documents corresponding to the multiple interaction rounds and the document reference data into a fourth intelligent agent, so that the fourth intelligent agent aggregates the multiple initial documents and the document reference data to generate the target document.

[0120] The fourth agent refers to an agent built based on a large language model. Specifically, the fourth agent can be an existing large language model, or can be obtained by fine-tuning the existing large language model, which is not limited in the present embodiment of the application.

[0121] By inputting multiple initial documents and document reference data into the fourth agent, the fourth agent can summarize the multiple initial documents, for example, perform deduplication processing on the document contents of the multiple initial documents, and then merge the deduplication document contents. For another example, further content analysis is performed on the multiple initial documents, so as to generate a summary document based on the multiple initial documents.

[0122] In addition, after aggregating multiple initial documents, it is necessary to fill the document reference data into the document section corresponding to the reference data type according to the reference data type corresponding to the document reference data, and finally generate the target document.

[0123] In an optional implementation, in order to improve the generation efficiency and generation effect of data processing tasks, the data processing operations corresponding to the associated data and the data processing examples corresponding to the data processing operations can be determined, thereby generating data processing tasks based on the data processing operations and the data processing examples.

[0124] Correspondingly, a data processing task is generated according to the associated data corresponding to the associated subject, including: determining the operation type of the data processing operation corresponding to the data type according to the data type corresponding to the associated data; performing the data processing operation on the preset data according to the operation type to obtain the type processing result of the preset data, and generating a data processing example corresponding to the data processing operation according to the type processing result; filling the associated data, the data processing operation corresponding to the associated data, and the data processing example into the preset task instruction template to generate the data processing task.

[0125] The specific explanation of the data type corresponding to the associated data and the operation type of the data processing operation can be referred to the above description, which will not be repeated here.

[0126] It should be noted that when there are multiple operation types of data processing operations, when determining the operation type of the data processing operation corresponding to the data type, one or more operation types can be selected according to actual data processing needs to perform the data processing operation corresponding to the operation type on the associated data. The embodiments of the present application do not limit this.

[0127] The preset data is exemplary data of the same data type as the associated data. Thus, according to the operation type, a data processing operation is performed on the preset data to obtain a type processing result, which is the data processing result of the data processing operation of the operation type performed on the preset data.

[0128] Thus, according to the type processing result of the preset data, a data processing example corresponding to the data processing operation can be obtained.

[0129] The preset task instruction template refers to a task instruction structure to be filled in. The preset task instruction template includes preset task description information, associated data information to be filled in, data processing operation information to be filled in, and data processing example information to be filled in.

[0130] Finally, a data processing task can be generated by filling the associated data, the data processing operations corresponding to the associated data, and the data processing examples into the corresponding positions of the preset task instruction template.

[0131] In an embodiment of the present application, by determining the data processing operation corresponding to the associated data and the data processing example corresponding to the data processing operation, the data processing operation, the associated data and the data processing example are filled into the preset task instruction template, thereby improving the task generation flexibility of the data processing task.

[0132] In an optional implementation, in order to achieve targeted data processing of different related data and improve data processing effects, the text data type and the time series data type can correspond to different operation types of data processing operations, respectively.

[0133] Correspondingly, in the case where the data type is a text data type, the operation types of the data processing operation corresponding to the data type include: text filtering type, text analysis type, and text aggregation type;

[0134] When the data type is a time series data type, the operation types of the data processing operation corresponding to the data type include: data statistics type, data prediction type, and data trend analysis type.

[0135] Among them, when the operation type is a text filtering type, the corresponding data processing operation needs to filter the associated data of the text data type. When the operation type is a text analysis type, the corresponding data processing operation needs to analyze and process the associated data of the text data type. When the operation type is a text aggregation type, the corresponding data processing operation needs to aggregate the associated data of the text data type.

[0136] Among them, when the operation type is data statistics type, the corresponding data processing operation needs to perform data statistics on the associated data of the time series data type. For example, calculate the maximum value, minimum value, average value of the data, calculate the preset indicator value corresponding to the data, such as the 7-day stock price average, etc.

[0137] When the operation type is data prediction type, the corresponding data processing operation needs to predict the associated data of the time series data type, for example, predict the data value and data change in the future.

[0138] When the operation type is data trend analysis type, the corresponding data processing operation needs to perform trend analysis on the associated data of the time series data type. For example, based on the double moving average strategy, analyze the rising and falling trends of stock data.

[0139] It should be noted that, since the text data type can be further divided according to data attributes, in order to achieve targeted processing of associated data of different data types, different operation types can correspond to text data types with different data attributes.

[0140] In an optional implementation, when the text data type is a news data type, the text filtering types include: keyword filtering type, time filtering type, similar text filtering type; the text analysis types include: topic analysis type, property analysis type, association analysis type; the text summary types include: topic summary type, time summary type, hierarchical summary type.

[0141] Among them, the keyword filtering type is used to represent the text filtering processing of the associated data of the news data type according to the preset keywords. The time filtering type is used to represent the text filtering processing of the associated data of the news data type according to the preset time node. The similar text filtering type is used to represent the content comparison of the associated data of the news data type, so as to filter the news information with similar content.

[0142] Among them, the topic analysis type is used to characterize the topic extraction of the associated data of the news data type. The property analysis type is used to characterize the content property of the associated data of the news data type, and obtain the property type corresponding to the associated data of the news data type, such as positive news or negative news. The association analysis type is used to characterize the association analysis of the associated data of the news data type, so as to obtain the common content of the associated data of the news data type.

[0143] Among them, the topic aggregation type is used to represent the aggregation of related data of news data types corresponding to similar topics. The time aggregation type is used to represent the aggregation of related data of news data types corresponding to preset time. The hierarchical aggregation type is used to represent the aggregation of related data of news data types according to preset hierarchical structure.

[0144] In an optional implementation, when the text data type is a document data type, the text filtering types include: format filtering type, topic filtering type; the text analysis types include: structure analysis type, content analysis type, aggregation analysis type; the text summary types include: content summary type, data summary type.

[0145] Among them, the format filtering type is used to represent the filtering of the associated data of the document data type according to the preset document format. For example, only the relevant document data of the report type is extracted. The subject filtering type is used to represent the filtering of the associated data of the document data type according to the preset subject, for example, only the sales analysis data of a certain product is extracted from the associated data.

[0146] Among them, the structural analysis type is used to represent the structural and formal analysis of the associated data of the document data type, and to extract the corresponding document structure. The content analysis type is used to represent the content analysis of the associated data of the document data type, and to obtain the key content information contained in the associated data. The aggregation analysis type is used to represent the aggregation analysis of the associated data of the document data type, and to mine the data logical relationship of the associated data.

[0147] The content summary type is used to indicate that the associated data of the document data type is summarized, for example, the key content information contained in the associated data is summarized. The data summary type is used to indicate that only the digital data content contained in the associated data of the document data type is summarized.

[0148] In an optional implementation, when the text data type is a basic data type, the text filtering type includes: dimension filtering type, indicator filtering type; the text analysis type includes: comparative analysis type, factor analysis type; the text summary type includes: summary type.

[0149] Among them, the dimension filtering type is used to represent the filtering of the associated data of the fundamental data type according to the preset data dimension, such as region, time, etc. The indicator filtering type is used to represent the filtering of the associated data of the fundamental data type according to the preset indicator, such as the preset financial indicator.

[0150] Among them, the comparative analysis type is used to characterize the comparative analysis of the associated data of the fundamental data type, for example, comparing the associated data of different time periods. The factor analysis type is used to characterize the analysis of the changes of preset factors from the associated data of the fundamental data type, for example, the changes in revenue.

[0151] The summary type is used to represent summarizing the associated data of the fundamental data type, that is, analyzing and summarizing the key information in the associated data to obtain summary text content.

[0152] In an optional implementation, in order to improve the data generation efficiency and data generation effect of document reference data, corresponding data processing tasks can be generated for the associated data corresponding to each data type, so that by setting a first intelligent agent for executing preset operations corresponding to different data types, multiple first intelligent agents can process multiple data processing tasks in parallel.

[0153] Correspondingly, the associated data corresponding to the associated subject includes: multiple associated data corresponding to multiple data types, the data processing task includes: multiple data processing tasks corresponding to the multiple associated data respectively, the number of first intelligent agents is multiple, and different first intelligent agents are used to perform preset operations corresponding to different data types. Then, the data processing task is assigned to the first preset intelligent agent, including: for any data processing task, according to the data type of the associated data corresponding to the data processing task, determine the first intelligent agent corresponding to the data type, and assign any data processing task to the first intelligent agent corresponding to the data type.

[0154] Among them, the preset operation corresponding to the data type is used to represent the operation that the first agent can perform on the associated data corresponding to the data type, which is determined according to the data processing operation corresponding to the data type.

[0155] Thus, for each data type of associated data, a corresponding dedicated first agent can be set for the data type. When allocating data processing tasks, the data processing task of the associated data corresponding to the data type can be directly allocated to the first agent corresponding to the data type.

[0156] Correspondingly, multiple first intelligent agents can process multiple data processing tasks in parallel at the same time, which not only effectively improves the data generation efficiency of the document reference data, but also, because each first intelligent agent is only dedicated to processing preset operations corresponding to one data type, the data quality of the generated document reference data is also effectively improved.

[0157] For ease of understanding, the technical implementation details in the above embodiment are described in detail below using a specific example:

[0158] The current way of generating research reports in the financial field is mainly based on the analysis and summary of experts in the financial field. The experts conduct analysis and writing from different angles based on their years of experience and unique understanding of market conditions. However, the expert-based writing method is difficult to generate research reports on a large scale and automatically, which is time-consuming and laborious.

[0159] Recent advances in large language models (LLMs) represent a major leap forward in the field of artificial intelligence. Cutting-edge models such as ChatGPT, Gemini, and LLaMA-3 have demonstrated promising reasoning capabilities, approaching average human performance in many areas. At the same time, these large language models extend traditional chat and text-based applications to more dynamic and proactive role scenarios, where they can not only generate text, but also autonomously plan and analyze in a variety of dialogue scenarios such as code writing and software design, device control, and network applications.

[0160] Of course, there is also a way to generate research reports based on large language models through research report data collected in the financial field. However, the existing method is to write directly through model fine-tuning or training. The model can only imitate the format of existing research reports and cannot guarantee the generation quality and flexibility.

[0161] In view of this, the embodiments of the present application provide a document generation method, a document generation device, an electronic device and a computer-readable medium, which realize the analysis and writing of research reports through multi-agent interactive learning.

[0162] Figure 2 A flowchart of a document generation method provided in an embodiment of the present application, referring to Figure 2 , the method comprising:

[0163] Step S201: determining an associated subject corresponding to a target document to be generated, and generating a plurality of data processing tasks corresponding to the plurality of associated data according to the plurality of associated data corresponding to the associated subject and the plurality of data types.

[0164] The document generation request can be parsed to determine the associated subject corresponding to the target document to be generated. For example, a user generates a document generation request by asking a question such as "how is the performance of a certain stock". Thus, by parsing the document generation request, it is determined that the associated subject is a certain stock.

[0165] Thus, a data processing task is generated according to the associated data corresponding to the associated subject.

[0166] It should be noted that the associated data may include associated data corresponding to multiple data types, and a corresponding data processing task may be generated for the associated data corresponding to each data type.

[0167] It should also be noted that this step can be performed by a preset central agent. Figure 3 A schematic diagram of a document generation method provided in an embodiment of the present application is shown, referring to Figure 3 .

[0168] The user makes a document generation request based on the target document to be generated, such as "how is the market today" or "how is the performance of a certain stock". After receiving the document generation request, the system forwards it to the central agent for processing.

[0169] The central agent receives input from the user, understands the current user's writing problem, analyzes the nature of the problem, and decomposes the problem, thereby identifying the associated subject corresponding to the target document to be generated, that is, extracting the key parameters in the problem. For example, is it to analyze and write about the entire A-share market, or to generate an analytical report on the data of a certain stock.

[0170] When parsing a document to generate a request, the associated intent corresponding to the associated subject can also be determined, that is, the intent corresponding to the question, such as stock market analysis.

[0171] Thus, different data processing tasks are ultimately generated according to the associated data of different data types corresponding to the associated subjects.

[0172] Step S202: For any data processing task, determine the first agent corresponding to the data type according to the data type of the associated data corresponding to the data processing task, and assign any data processing task to the first agent corresponding to the data type, so that the first agent processes the associated data according to the data processing task to generate document reference data.

[0173] For example, continue to refer to Figure 3 , the central intelligent agent can call multiple first intelligent agents for joint analysis, that is, assign data processing tasks to the corresponding first intelligent agents.

[0174] Accordingly, the first agent may include:

[0175] News analysis agent: It is used to process the associated data corresponding to the news data type, that is, to perform data processing tasks corresponding to the associated data. For example, through the associated subject identified by the central agent, that is, the question parameter, it obtains and analyzes the latest associated data such as market news, and performs information filtering and summary.

[0176] Industry research report analysis agent: It is used to process the associated data corresponding to the document data type, that is, to perform the data processing tasks corresponding to the associated data. For example, through the associated subject identified by the central agent, analyze and organize the existing public industry research reports related to the current stock.

[0177] Time series signal analysis agent: It is used to process the associated data corresponding to the time series data type, that is, to execute the data processing tasks corresponding to the associated data. For example, based on the associated subject identified by the central agent, the time series data of the current market index or individual stocks in the recent period are processed and analyzed, and the technical analysis report is extracted, including but not limited to the 7-day stock price average and the 14-day stock price average. Based on the double moving average strategy, it is analyzed whether the individual stock has an upward or downward trend.

[0178] The company's public fundamental data analysis agent: It is used to process the associated data corresponding to the fundamental data type, that is, to perform the data processing tasks corresponding to the associated data. For example, through the associated entities identified by the central agent, the company's announcements and related financial and other fundamental operating data are analyzed, and key information is extracted for analysis and summary.

[0179] Other types of agents: They are used to process the associated data corresponding to other preset data types, that is, to perform data processing tasks corresponding to the associated data. For example, a corresponding first agent can be constructed for a company's internal unique data source, so as to process some internal non-public research report analysis data.

[0180] Thus, each first agent extracts relevant information and performs corresponding data processing tasks according to its division of labor.

[0181] Step S203: Input the document reference data into multiple second agents corresponding to different preset roles, wherein different preset roles correspond to different evaluation dimensions, and the multiple second agents are set in the same group chat environment, so that the multiple second agents perform multiple rounds of interactive evaluation on the document reference data according to the evaluation dimensions corresponding to the second agents in the group chat environment, and obtain multiple rounds of interactive data between the multiple second agents.

[0182] For example, continue to refer to Figure 3 , where the second agent is the KOL Agents (key opinion leader agent) in the figure.

[0183] The document reference data generated by the first agent is summarized and forwarded to the second agent, namely KOL Agents (k1 to kN). Thus, the KOL Agents evaluate the document reference data from different evaluation dimensions by simulating the analysis and opinions of different preset roles. Thus, these KOL Agents interact with each other in a group chat environment for multiple rounds to reach a consensus or put forward different opinions.

[0184] Step S204: extracting data from each round of interaction based on the third agent in the group chat environment to obtain key interaction data corresponding to multiple second agents in each interaction round.

[0185] For example, continue to refer to Figure 3 ,In the group chat environment, the analysis and discussion results of the KOL Agents will be displayed and further discussed. ,A third agent is also set up in the group chat environment, namely the ,observer agent in the figure, which extracts key interaction data, i.e. valuable opinion content, by supervising the discussion process between the second agents.

[0186] Step S205: For each interaction round, generate an evaluation result corresponding to each interaction round according to the key interaction data corresponding to the plurality of second agents.

[0187] Step S206: Obtain the evaluation result corresponding to the document reference data according to the evaluation result corresponding to each interaction round.

[0188] Step S207: Generate a target document according to the evaluation result of the document reference data.

[0189] The evaluation results of the document reference data include evaluation results of multiple document reference data corresponding to multiple reference data types in multiple interaction rounds. Thus, for each interaction round, the third agent can generate an initial document corresponding to each interaction round.

[0190] Furthermore, a fourth agent may be provided to generate a final target document based on initial documents corresponding to multiple interaction rounds.

[0191] For example, continue to refer to Figure 3 , the fourth agent, the summary agent, summarizes the initial documents and document reference data corresponding to each interaction round. In multiple rounds of dialogue, the second agents will interact with each other and iteratively optimize the content of their respective opinions. Finally, the fourth agent conducts overall analysis and understanding, outputs the summarized document content, and writes the final target document.

[0192] It should also be noted that the performance of each agent can be evaluated through the reward mechanism Reward. Based on the evaluation results of the document reference data obtained from the dialogue interaction, the reward mechanism can correspond to multiple standards, such as analyzing the accuracy of the document reference data and the completeness of the evaluation results. The feedback-based reward mechanism provides feedback guidance to the generation results of the agent, so that the agent can further produce more complete and detailed content. The reward mechanism helps to optimize the performance of the agent and promote the continuous improvement of the entire system.

[0193] In an embodiment of the present application, the analysis and writing of research report documents are realized based on a multi-agent interactive learning method, and research report documents can be generated on a large scale. In addition, the second agents corresponding to multiple preset roles have different evaluation dimensions, thereby being able to generate unique insights under the evaluation dimensions. The embodiment of the present application realizes the improvement of viewpoints in the research report content through multi-agent group chat dialogues, and generates comprehensive research report documents.

[0194] It can be understood that the above-mentioned various method embodiments mentioned in this application can be combined with each other to form a combined embodiment without violating the principle logic. Due to space limitations, this application will not go into details. Those skilled in the art can understand that in the above-mentioned method of the specific implementation method, the specific execution order of each step should be determined by its function and possible internal logic.

[0195] In addition, the present application also provides a document generation device, an electronic device, and a computer-readable storage medium, all of which can be used to implement any document generation method provided by the present application. The corresponding technical solutions and descriptions are referred to in the corresponding records of the method part and will not be repeated here.

[0196] Figure 4 A block diagram of a document generation device provided in an embodiment of the present application.

[0197] Reference Figure 4 , the embodiment of the present application provides a document generation device, the document generation device 40 includes:

[0198] The task construction module 41 is used to determine the associated subject corresponding to the target document to be generated, and generate a data processing task according to the associated data corresponding to the associated subject;

[0199] A data generation module 42, configured to assign the data processing task to a preset first agent, so that the first agent processes the associated data according to the data processing task to generate document reference data;

[0200] A data evaluation module 43, used for inputting the document reference data into a preset second agent, so that the second agent evaluates the document reference data to obtain an evaluation result of the document reference data;

[0201] The document generation module 44 is used to generate the target document according to the evaluation result of the document reference data.

[0202] In an optional implementation, the preset second agent includes: a plurality of second agents corresponding to different preset roles, and different preset roles correspond to different evaluation dimensions; and the plurality of second agents are set in the same group chat environment, and the inputting of the document reference data into the preset second agent so that the second agent evaluates the document reference data, and obtaining the evaluation result of the document reference data includes:

[0203] Inputting the document reference data into the plurality of second agents, so that the plurality of second agents evaluate the document reference data in a plurality of interaction rounds according to the evaluation dimensions corresponding to the second agents in the group chat environment, and obtaining a plurality of rounds of interaction data between the plurality of second agents;

[0204] An evaluation result corresponding to the document reference data is obtained according to the multiple rounds of interaction data.

[0205] In an optional implementation, a third agent is provided in the group chat environment, and obtaining the evaluation result corresponding to the document reference data according to the multiple rounds of interaction data includes:

[0206] Extracting data from each round of interaction data based on the third agent to obtain key interaction data corresponding to multiple second agents in each interaction round;

[0207] For each interaction round, generating an evaluation result corresponding to each interaction round according to the key interaction data corresponding to the plurality of second agents;

[0208] According to the evaluation results corresponding to each interaction round, the evaluation results corresponding to the document reference data are obtained.

[0209] In an optional implementation, the evaluation results of the document reference data include evaluation results of multiple document reference data corresponding to multiple reference data types in multiple interaction rounds, and generating a target document according to the evaluation results of the document reference data includes:

[0210] Determining, according to the multiple reference data types corresponding to the multiple document reference data, multiple document sections corresponding to the multiple reference data types;

[0211] Generating a document template corresponding to the target document to be generated according to the multiple document sections;

[0212] For evaluation results of multiple document reference data corresponding to multiple reference data types in each interaction round, fill the evaluation results of the document reference data corresponding to each reference data type into the document section corresponding to the reference data type in the document template to generate an initial document corresponding to each interaction round;

[0213] The target document is generated according to a plurality of initial documents corresponding to a plurality of interaction rounds.

[0214] In an optional implementation, generating a data processing task according to the associated data corresponding to the associated subject includes:

[0215] Determining, according to a data type corresponding to the associated data, an operation type of a data processing operation corresponding to the data type;

[0216] According to the operation type, a data processing operation is performed on the preset data to obtain a type processing result of the preset data, and according to the type processing result, a data processing example corresponding to the data processing operation is generated;

[0217] The associated data, the data processing operations corresponding to the associated data, and the data processing examples are filled into a preset task instruction template to generate the data processing task.

[0218] In an optional implementation, the data type includes a text data type and a time series data type;

[0219] In the case where the data type is a text data type, the operation type of the data processing operation corresponding to the data type includes: a text filtering type, a text analysis type, and a text aggregation type;

[0220] In the case where the data type is a time series data type, the operation type of the data processing operation corresponding to the data type includes: a data statistics type, a data prediction type, and a data trend analysis type.

[0221] In an optional implementation, the associated data corresponding to the associated subject includes: multiple associated data corresponding to multiple data types, the data processing task includes: multiple data processing tasks corresponding to the multiple associated data respectively, the number of the first agents is multiple, and different first agents are used to perform preset operations corresponding to different data types, then the data processing task is assigned to the first preset agent, including:

[0222] For any data processing task, determine the first agent corresponding to the data type according to the data type of the associated data corresponding to the data processing task, and assign the any data processing task to the first agent corresponding to the data type.

[0223] Each module in the document generation device can be implemented in whole or in part by software, hardware, or a combination thereof. Each module can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in a computer device in the form of software, so that the processor can call and execute operations corresponding to each module.

[0224] Figure 5 A block diagram of an electronic device provided in an embodiment of the present application.

[0225] Reference Figure 5 An embodiment of the present application provides an electronic device, which includes: at least one processor 501; at least one memory 502, and one or more I / O interfaces 503, connected between the processor 501 and the memory 502; wherein the memory 502 stores one or more computer programs that can be executed by at least one processor 501, and the one or more computer programs are executed by at least one processor 501 so that the at least one processor 501 can execute the above-mentioned document generation method.

[0226] Each module in the above electronic device can be implemented in whole or in part by software, hardware, or a combination thereof. Each module can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in a computer device in the form of software, so that the processor can call and execute operations corresponding to each module.

[0227] The embodiment of the present application also provides a computer-readable storage medium on which a computer program is stored, wherein the computer program implements the above-mentioned document generation method when executed by a processor / processing core. The computer-readable storage medium may be a volatile or non-volatile computer-readable storage medium.

[0228] An embodiment of the present application also provides a computer program product, including a computer-readable code, or a non-volatile computer-readable storage medium carrying the computer-readable code. When the computer-readable code runs in a processor of an electronic device, the processor in the electronic device executes the above-mentioned document generation method.

[0229] It will be appreciated by those skilled in the art that all or some of the steps, systems, and functional modules / units in the methods disclosed above may be implemented as software, firmware, hardware, and appropriate combinations thereof. In a hardware implementation, the division between the functional modules / units mentioned in the above description does not necessarily correspond to the division of physical components; for example, a physical component may have multiple functions, or a function or step may be performed by several physical components in cooperation. Some or all physical components may be implemented as software executed by a processor, such as a central processing unit, a digital signal processor, or a microprocessor, or may be implemented as hardware, or may be implemented as an integrated circuit, such as an application-specific integrated circuit. Such software may be distributed on a computer-readable storage medium, which may include a computer storage medium (or a non-transitory medium) and a communication medium (or a temporary medium).

[0230] As is known to those of ordinary skill in the art, the term computer storage media includes volatile and nonvolatile, removable and non-removable media implemented in any method or technology for storing information (such as computer-readable program instructions, data structures, program modules or other data). Computer storage media include, but are not limited to, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), static random access memory (SRAM), flash memory or other memory technology, portable compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical disk storage, magnetic cassettes, magnetic tapes, magnetic disk storage or other magnetic storage devices, or any other medium that can be used to store desired information and can be accessed by a computer. In addition, it is known to those of ordinary skill in the art that communication media typically contain computer-readable program instructions, data structures, program modules or other data in a modulated data signal such as a carrier wave or other transmission mechanism, and may include any information delivery medium.

[0231] The computer-readable program instructions described herein can be downloaded from a computer-readable storage medium to each computing / processing device, or downloaded to an external computer or external storage device via a network, such as the Internet, a local area network, a wide area network, and / or a wireless network. The network can include copper transmission cables, optical fiber transmissions, wireless transmissions, routers, firewalls, switches, gateway computers, and / or edge servers. The network adapter card or network interface in each computing / processing device receives the computer-readable program instructions from the network and forwards the computer-readable program instructions for storage in the computer-readable storage medium in each computing / processing device.

[0232] The computer program instructions for performing the operation of the present application can be assembler instructions, instruction set architecture (ISA) instructions, machine instructions, machine-related instructions, microcode, firmware instructions, state setting data, or source code or object code written in any combination of one or more programming languages, including object-oriented programming languages, such as Smalltalk, C++, etc., and conventional procedural programming languages, such as "C" language or similar programming languages. Computer-readable program instructions can be executed completely on a user's computer, partially on a user's computer, executed as an independent software package, partially on a user's computer, partially on a remote computer, or completely on a remote computer or server. In the case of a remote computer, the remote computer can be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or can be connected to an external computer (for example, using an Internet service provider to connect through the Internet). In some embodiments, by using the state information of a computer-readable program instruction to personalize an electronic circuit, such as a programmable logic circuit, a field programmable gate array (FPGA) or a programmable logic array (PLA), the electronic circuit can execute a computer-readable program instruction, thereby realizing various aspects of the present application.

[0233] The computer program product described herein may be implemented in hardware, software, or a combination thereof. In one optional embodiment, the computer program product is embodied as a computer storage medium, and in another optional embodiment, the computer program product is embodied as a software product, such as a software development kit (SDK), etc.

[0234] Various aspects of the present application are described herein with reference to the flowcharts and / or block diagrams of the methods, devices (systems) and computer program products according to the embodiments of the present application. It should be understood that each box in the flowchart and / or block diagram and the combination of each box in the flowchart and / or block diagram can be implemented by computer-readable program instructions.

[0235] These computer-readable program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, thereby producing a machine, so that when these instructions are executed by the processor of the computer or other programmable data processing device, a device that implements the functions / actions specified in one or more boxes in the flowchart and / or block diagram is generated. These computer-readable program instructions can also be stored in a computer-readable storage medium, and these instructions cause the computer, programmable data processing device, and / or other equipment to work in a specific manner, so that the computer-readable medium storing the instructions includes a manufactured product, which includes instructions for implementing various aspects of the functions / actions specified in one or more boxes in the flowchart and / or block diagram.

[0236] Computer-readable program instructions may also be loaded onto a computer, other programmable data processing apparatus, or other device so that a series of operating steps are performed on the computer, other programmable data processing apparatus, or other device to produce a computer-implemented process, thereby causing the instructions executed on the computer, other programmable data processing apparatus, or other device to implement the functions / actions specified in one or more boxes in the flowchart and / or block diagram.

[0237] The flow chart and block diagram in the accompanying drawings show the possible architecture, function and operation of the system, method and computer program product according to multiple embodiments of the present application. In this regard, each square box in the flow chart or block diagram can represent a part of a module, program segment or instruction, and a part of the module, program segment or instruction includes one or more executable instructions for realizing the logical function of the specification. In some alternative implementations, the function marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two continuous square boxes can actually be executed substantially in parallel, and they can also be executed in the opposite order sometimes, depending on the function involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be realized by a dedicated hardware-based system that performs the function or action of the specification, or can be realized by a combination of special-purpose hardware and computer instructions.

[0238] Example embodiments have been disclosed herein, and although specific terms are employed, they are used and should be interpreted only in a general illustrative sense and not for limiting purposes. In some instances, it will be apparent to those skilled in the art that, unless otherwise expressly noted, features, characteristics, and / or elements described in conjunction with a particular embodiment may be used alone or in combination with features, characteristics, and / or elements described in conjunction with other embodiments. Therefore, those skilled in the art will appreciate that various changes in form and detail may be made without departing from the scope of the present application as set forth in the appended claims.

Claims

1. A document generation method, characterized in that: The method comprises: Determine the associated subject corresponding to the target document to be generated, and generate a data processing task according to the associated data corresponding to the associated subject; Allocating the data processing task to a preset first agent, so that the first agent processes the associated data according to the data processing task to generate document reference data; Inputting the document reference data into a preset second agent, so that the second agent evaluates the document reference data to obtain an evaluation result of the document reference data; The target document is generated according to the evaluation result of the document reference data.

2. The method according to claim 1, characterized in that The preset second agent includes: a plurality of second agents corresponding to different preset roles, and different preset roles correspond to different evaluation dimensions; and the plurality of second agents are set in the same group chat environment, and the document reference data is input into the preset second agent so that the second agent evaluates the document reference data, and the evaluation result of the document reference data is obtained, including: Inputting the document reference data into the plurality of second agents, so that the plurality of second agents evaluate the document reference data in a plurality of interaction rounds according to the evaluation dimensions corresponding to the second agents in the group chat environment, and obtaining a plurality of rounds of interaction data between the plurality of second agents; An evaluation result corresponding to the document reference data is obtained according to the multiple rounds of interaction data.

3. The method according to claim 2, characterized in that A third agent is provided in the group chat environment, and obtaining the evaluation result corresponding to the document reference data according to the multiple rounds of interaction data includes: Extracting data from each round of interaction data based on the third agent to obtain key interaction data corresponding to multiple second agents in each interaction round; For each interaction round, generating an evaluation result corresponding to each interaction round according to the key interaction data corresponding to the plurality of second agents; According to the evaluation results corresponding to each interaction round, the evaluation results corresponding to the document reference data are obtained.

4. The method according to any one of claims 2 to 3, characterized in that: The evaluation results of the document reference data include evaluation results of multiple document reference data corresponding to multiple reference data types in multiple interaction rounds, and generating a target document according to the evaluation results of the document reference data includes: Determining, according to the multiple reference data types corresponding to the multiple document reference data, multiple document sections corresponding to the multiple reference data types; Generating a document template corresponding to the target document to be generated according to the multiple document sections; For evaluation results of multiple document reference data corresponding to multiple reference data types in each interaction round, fill the evaluation results of the document reference data corresponding to each reference data type into the document section corresponding to the reference data type in the document template to generate an initial document corresponding to each interaction round; The target document is generated according to a plurality of initial documents corresponding to a plurality of interaction rounds.

5. The method according to claim 1, characterized in that The step of generating a data processing task according to the associated data corresponding to the associated subject includes: Determining, according to a data type corresponding to the associated data, an operation type of a data processing operation corresponding to the data type; According to the operation type, a data processing operation is performed on the preset data to obtain a type processing result of the preset data, and according to the type processing result, a data processing example corresponding to the data processing operation is generated; The associated data, the data processing operations corresponding to the associated data, and the data processing examples are filled into a preset task instruction template to generate the data processing task.

6. The method according to claim 5, characterized in that The data types include text data types and time series data types; In the case where the data type is a text data type, the operation type of the data processing operation corresponding to the data type includes: a text filtering type, a text analysis type, and a text aggregation type; In the case where the data type is a time series data type, the operation type of the data processing operation corresponding to the data type includes: a data statistics type, a data prediction type, and a data trend analysis type.

7. The method according to claim 5, characterized in that The associated data corresponding to the associated subject includes: multiple associated data corresponding to multiple data types, the data processing task includes: multiple data processing tasks corresponding to the multiple associated data respectively, the number of the first intelligent agents is multiple, and different first intelligent agents are used to perform preset operations corresponding to different data types, then the data processing task is assigned to the first preset intelligent agent, including: For any data processing task, determine the first agent corresponding to the data type according to the data type of the associated data corresponding to the data processing task, and assign the any data processing task to the first agent corresponding to the data type.

8. A document generation device, characterized in that: The device comprises: A task construction module, used to determine the associated subject corresponding to the target document to be generated, and generate a data processing task according to the associated data corresponding to the associated subject; A data generation module, used for allocating the data processing task to a preset first agent, so that the first agent processes the associated data according to the data processing task to generate document reference data; A data evaluation module, used for inputting the document reference data into a preset second agent, so that the second agent evaluates the document reference data to obtain an evaluation result of the document reference data; The document generation module is used to generate the target document according to the evaluation result of the document reference data.

9. An electronic device, characterized in that: include: at least one processor; as well as a memory communicatively connected to the at least one processor; wherein, The memory stores one or more computer programs executable by the at least one processor, and the one or more computer programs are executed by the at least one processor to enable the at least one processor to execute the document generation method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the computer program implements the document generation method according to any one of claims 1 to 7.

Images