Personalized research report generation method and related device

Abstract

The invention discloses a personalized research report generation method and a related device, and relates to the technical field of data processing. Comprising the steps of obtaining reading habit features, content preference features and analysis dimension preference features of a target user; obtaining the preprocessed research report text; generating structured research report information through a large language model, the preprocessed research report text, the user demand portrait and the first cue word; based on the reading habit features and the content preference features, determining a research report template; and through the large language model, the research report template and the structured research report information, analyzing the dimension preference features and the second cue word, and generating a personalized research report. According to the method, the research report generation logic is customized based on the multi-dimensional demand characteristics of the user, structured processing and personalized output of the research report are completed based on the large language model, and the personalized research report matched with the user demand can be efficiently obtained.

Description

Technical Field

[0001] This application relates to the field of data processing technology, and in particular to a method and apparatus for generating personalized research reports. Background Technology

[0002] Research reports are in-depth research reports compiled on the development trends of a target industry, the operating conditions of a company, or the trends of the financial market. Analyzing research reports can uncover hidden market patterns, identify investment opportunities, and avoid potential investment risks.

[0003] Existing research report analysis methods involve researchers reading a large number of research reports and manually filtering key information, which suffers from low information acquisition efficiency, information overload, and low matching degree to personalized needs.

[0004] Therefore, how to efficiently obtain personalized research reports that match user needs has become one of the urgent technical problems to be solved in the fields of financial investment research and data processing. Summary of the Invention

[0005] To address the aforementioned issues, this application provides a personalized research report generation method for efficiently obtaining personalized research reports that match user needs.

[0006] The embodiments of this application disclose the following technical solutions:

[0007] The first aspect of this application provides a method for generating personalized research reports, which includes:

[0008] Obtain a user demand profile of the target users; the user demand profile includes reading habit characteristics, content preference characteristics, and analysis dimension preference characteristics.

[0009] Obtain the preprocessed research report text;

[0010] Structured research report information is generated using a large language model, the preprocessed research report text, the user demand profile, and the first prompt word.

[0011] Based on the aforementioned reading habit characteristics and content preference characteristics, a research report template is determined;

[0012] Personalized research reports are generated using the large language model, the research report template, the structured research report information, the analytical dimension preference features, and the second prompt words.

[0013] In one optional implementation, the user demand profile further includes work rhythm characteristics, and the method further includes:

[0014] Based on the work rhythm characteristics and the timeliness of the information in the personalized research report, the time period for pushing the research report is determined;

[0015] During the specified research report push period, the personalized research report is pushed to the target user.

[0016] In one optional implementation, obtaining the user demand profile of the target user includes:

[0017] Obtain the demand questionnaire data filled out by the target users; the demand questionnaire data includes the industry sub-sectors that users are interested in, the enterprise operation indicators that users are interested in, the users' reading habits, and the users' work rhythm;

[0018] Collect reading behavior data of the target users when reading research reports;

[0019] Extract information search behavior data of the target user within a preset time period;

[0020] Based on the demand questionnaire data, the reading behavior data, and the information search behavior data, the user demand profile is drawn.

[0021] In one optional implementation, obtaining the preprocessed research report text includes:

[0022] Data from the first research report is collected via API interface at a preset data collection frequency.

[0023] The first research report data is processed according to a preset research report format to generate the second research report data; the preset research report format includes arranging the data in the order of research report number, publication time, industry, company, and full text.

[0024] The second research report data is parsed and annotated to obtain the preprocessed research report text.

[0025] In one optional implementation, determining the research report push time period based on the work rhythm characteristics and the timeliness of information in the personalized research report includes:

[0026] If the timeliness of the information in the personalized research report is high, then the target time period starting from the current time will be used as the time period for pushing the research report.

[0027] If the timeliness of the information in the personalized research report is at a medium level, the time period for pushing the research report is determined based on the work rhythm characteristics.

[0028] In one alternative implementation, the method further includes:

[0029] If the target user searches for research report data of the target company in real time, the timeliness of the information related to the target company in the personalized research report information will be modified to the higher level.

[0030] In one alternative implementation, it also includes:

[0031] Obtain feedback information from the target user regarding the personalized research report;

[0032] Based on the feedback information, update the user demand profile and the research report template.

[0033] The second aspect of this application provides a personalized research report generation device, comprising:

[0034] The demand profile determination module is used to create user demand profiles; the user demand profiles include reading habit characteristics, content preference characteristics, and analysis dimension preference characteristics.

[0035] The research report text determination module is used to obtain the preprocessed research report text;

[0036] The structured research report information generation module is used to generate structured research report information using a large language model, the preprocessed research report text, the user demand profile, and the first prompt word.

[0037] The research report template determination module is used to determine the research report template based on the reading habit characteristics and the content preference characteristics;

[0038] The research report generation module is used to generate personalized research reports using the large language model, the research report template, the structured research report information, the analysis dimension preference features, and the second prompt words.

[0039] A third aspect of this application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method described in any implementation of the first aspect.

[0040] A fourth aspect of this application provides an electronic device, comprising:

[0041] A memory on which computer programs are stored;

[0042] A processor for executing the computer program in the memory to implement the steps of the method described in any implementation of the first aspect.

[0043] Compared with the prior art, this application has the following beneficial effects:

[0044] This application provides a personalized research report generation method, including: acquiring the reading habit characteristics, content preference characteristics, and analytical dimension preference characteristics of target users; acquiring preprocessed research report text; generating structured research report information through a large language model, the preprocessed research report text, user demand profiles, and first prompt words; determining a research report template based on reading habit characteristics and content preference characteristics; and generating a personalized research report through a large language model, the research report template, structured research report information, analytical dimension preference characteristics, and second prompt words. This application customizes the research report generation logic based on users' multi-dimensional demand characteristics and relies on a large language model to complete the structured processing and personalized output of the research report, enabling efficient acquisition of personalized research reports matching user needs. Attached Figure Description

[0045] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0046] Figure 1 A flowchart illustrating a personalized research report generation method provided in this application embodiment;

[0047] Figure 2 This is a schematic diagram of structured research report information provided in an embodiment of this application;

[0048] Figure 3 A schematic diagram illustrating a personalized research report provided in an embodiment of this application;

[0049] Figure 4 This is a schematic diagram of a personalized research report generation device provided in an embodiment of this application. Detailed Implementation

[0050] Taking a real-world investment research scenario as an example: Investment researcher A has long been deeply involved in the new energy industry, focusing on key indicators such as gross profit margin and R&D investment. Daily, they need to sift through a massive amount of research reports to find content that matches their needs. However, under traditional analysis methods, researchers must review the titles, abstracts, and even the full text of each report to locate the target information. The average time spent filtering each valid report exceeds 30 minutes, and key data is easily missed. Investment researcher B focuses on the consumer industry, paying close attention to core indicators such as market share and user repurchase rate. They also face the problems of low information filtering efficiency and poor matching of content with personalized needs. In summary, existing research report analysis methods suffer from technical pain points such as low information acquisition efficiency, information overload, and low matching with personalized needs.

[0051] Therefore, how to efficiently obtain personalized research reports that match user needs has become one of the urgent technical problems to be solved in the fields of financial investment research and data processing.

[0052] Based on this, this application provides a personalized research report generation method, including: obtaining the target user's reading habit characteristics, content preference characteristics, and analytical dimension preference characteristics; obtaining preprocessed research report text; generating structured research report information through a large language model, the preprocessed research report text, user demand profiles, and first prompt words; determining a research report template based on reading habit characteristics and content preference characteristics; and generating a personalized research report through a large language model, the research report template, structured research report information, analytical dimension preference characteristics, and second prompt words. This application customizes the research report generation logic based on the user's multi-dimensional demand characteristics and relies on a large language model to complete the structured processing and personalized output of the research report, enabling efficient acquisition of personalized research reports matching user needs.

[0053] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present application.

[0054] Figure 1 This is a flowchart illustrating a personalized research report generation method provided in an embodiment of this application. (Combined with...) Figure 1 As shown, the personalized research report generation method disclosed in this application includes:

[0055] S101, Obtain a user demand profile of the target user.

[0056] The user profile in this application includes reading habit characteristics, content preference characteristics, analysis dimension preference characteristics, and work rhythm characteristics.

[0057] Reading habit characteristics refer to users' behavioral preferences during the research report reading process, including the duration of research report reading, preferred reading media, and priority reading sections of research reports (such as abstracts, core data, conclusions, etc.).

[0058] Content preference characteristics refer to users' preferences for the topics, industries, companies, and content depth covered in research reports, including the target industry sectors, the types of companies they focus on, and the types of information they emphasize (such as policy interpretation, financial analysis, and market forecasts).

[0059] Analysis dimension preference features refer to the core judgment dimensions that users focus on during the research report analysis process, including preferred financial indicators (such as gross profit margin and revenue growth rate), industry indicators (such as market penetration rate and capacity utilization rate), and corporate operating indicators (such as R&D investment ratio and market share).

[0060] Work rhythm characteristics refer to the patterns in the time users use research reports based on their investment research work schedules, including the time periods when research reports are needed and the number of research reports obtained at one time.

[0061] In one optional implementation, the user demand profile of the target user is obtained, including:

[0062] A1, Obtain the questionnaire data filled out by the target user.

[0063] The data in the demand questionnaire includes industry sub-sectors that users are interested in, enterprise operation indicators that users are interested in, users' reading habits, and users' work rhythm.

[0064] Among them, users' reading habits include content structure preferences (indirect or detailed), data presentation preferences (table, chart, or text description), and content detail preferences (core data priority or logical deduction priority).

[0065] After obtaining the target users' needs questionnaire data, the questionnaire data can be stored in the "User Needs Database".

[0066] A2, Collect reading behavior data of the target users when reading research reports.

[0067] The reading behavior data in this application includes: key content identified based on reading time, core information extracted based on annotation operations, search keywords compiled based on search history, and user reading preference data determined based on collection or deletion operations.

[0068] For example, if a section of content takes more than 5 minutes to read, it will be marked as key content; highlighted and annotated content will be marked as core information; saved research reports will be marked as preferred content, and deleted research reports will be marked as irrelevant content.

[0069] After obtaining reading behavior data, it can be synchronized to the server hourly.

[0070] A3, extract the information search behavior data of the target user within a preset time period.

[0071] For example, if a target user frequently searches for "gross profit margin of a certain car company" within a preset time period (such as this week), that company will be listed as a key company to watch; if a target user frequently marks content related to "quarterly trend analysis", then "trend analysis" will be listed as a core content requirement; the key companies to watch and core content requirements obtained based on the above search behavior are the information search behavior data in this application.

[0072] A4. Based on the demand questionnaire data, the reading behavior data, and the information search behavior data, draw the user demand profile.

[0073] Specifically, the explicit demand characteristics in the demand questionnaire data are structured and analyzed to extract basic attributes such as the industries, indicators, reading habits, and work rhythms that users are interested in. At the same time, feature mining is performed on reading behavior data and information search behavior data to extract implicit demand characteristics such as users' key content focus, core information preferences, and search keyword preferences. Subsequently, the explicit and implicit demand characteristics are integrated to construct a user demand profile covering reading habit characteristics, content preference characteristics, analysis dimension preference characteristics, and work rhythm characteristics, and this profile is stored in the user demand graph of the user demand database.

[0074] It should be noted that this application can push a "demand verification questionnaire" to the client every month, and users can provide feedback on the matching degree between the current research report content and the demand (such as "complete match, partial match and mismatch") and optimization suggestions (such as "need to add competitor comparison"). The server adjusts the indicator weights, content preferences and other parameters in the demand graph based on the feedback to achieve continuous iteration of the demand model.

[0075] S102, Obtain the preprocessed research report text.

[0076] The preprocessed research report text can be obtained in the following ways:

[0077] B1, collects the first research report data through the API interface according to the preset data collection frequency.

[0078] For example, the latest research report data can be collected hourly from data sources corresponding to brokerage research report platforms, corporate financial reports, and industry indicator data via API interface, and used as the first research report data; the first research report data supports both PDF and Word formats.

[0079] B2. Process the first research report data according to the preset research report format to generate the second research report data.

[0080] The preset research report layout format includes arranging the reports in the following order: research report number, publication time, industry, company, and full text.

[0081] This step means organizing the first research report data according to the prescribed format requirements to obtain the second research report data; and storing the second research report data in the research report information database.

[0082] B2, perform text parsing and text annotation on the second research report data to obtain the preprocessed research report text.

[0083] By combining OCR recognition tools with a text parsing library, the second research report data in PDF and Word formats were converted into plain text. Then, non-core invalid information such as headers, footers, advertisements, and disclaimers were filtered and removed. Finally, the plain text content was segmented and annotated according to the logical structure of industry analysis, company analysis, financial data, and conclusions to obtain the preprocessed research report text.

[0084] Furthermore, the preprocessed research report text is stored in the research report information database.

[0085] S103, generate structured research report information using the large language model, the preprocessed research report text, the user demand profile, and the first prompt word.

[0086] After obtaining the preprocessed research report text and user demand profile, the above two items and the first prompt word are simultaneously input into the large language model (deployed on a GPU node), and the large language model outputs structured research report information.

[0087] For example, the first prompt could be: Based on the user demand map (industry: new energy vehicle companies, core indicators: gross profit margin, R&D investment, analysis dimension: quarterly trend), extracted from the following research report text:

[0088] (1) Gross profit margin data and year-on-year and quarter-on-quarter changes of new energy vehicle companies (which need to be matched with the list of companies followed by users) for the past three quarters;

[0089] (2) Amount of R&D investment and its percentage of revenue;

[0090] (3) Core factors affecting gross profit margin trends (such as raw material prices and sales volume);

[0091] (4) Industry comparison of relevant data (such as the difference in gross profit margin with leading car companies).

[0092] The extracted results should be output in the format of "Indicator Name-Data-Analysis-Source (Research Report Paragraph)," prioritizing the retention of core information and removing irrelevant content (such as data on upstream materials for new energy).

[0093] Understandably, the large language model performs semantic analysis on the preprocessed research report text based on the first prompt word, extracts information that matches the needs of the target users (prioritizing the extraction of information on companies that the target users are particularly interested in); and sorts the extracted information by importance, labeling the source of the information (research report number and paragraph information), thereby ensuring the traceability of data information.

[0094] Figure 2 This is a schematic diagram illustrating a structured research report information provided in an embodiment of this application. (In conjunction with...) Figure 2 As shown, the structured research report information includes user ID, research report ID, indicator name, data type, analysis item, data source, and importance ranking.

[0095] S104. Based on the reading habit characteristics and the content preference characteristics, determine the research report template.

[0096] For example, if user A's reading habits and content preferences are "concise structure, tabular data, and priority of core conclusions," then the first research report template will be matched and recommended to user A based on these characteristics. The content structure of the first research report template is "core conclusions, key data (tables), and risk warnings." This template highlights key information by presenting core conclusions first, presents key data intuitively in tabular form, and includes necessary risk warnings to meet user A's need for concise and efficient access to core information.

[0097] For example, if user B's reading habits and content preferences prioritize "detailed structure, charts, and logical deduction," then a second research report template will be matched and recommended to user B based on this characteristic. The content structure of this second research report template is "indicator trend analysis (charts), logical deduction, competitor comparison, and conclusion." This template displays indicator trend changes in chart form, accompanied by a complete logical deduction process and competitor comparison analysis, and finally outputs a conclusion, thus meeting user B's needs for in-depth analysis and rigorous logical deduction.

[0098] Furthermore, the research report templates corresponding to the target users are associated with the user IDs and stored in the system resource library.

[0099] S105, generate a personalized research report using the large language model, the research report template, the structured research report information, the analysis dimension preference features, and the second prompt words.

[0100] The research report template, structured research report information, user demand map analysis dimension preferences, and second prompt words are simultaneously input into the large language model, and the large language model outputs personalized research reports.

[0101] For example, the second prompt can be a personalized research report generated based on the following template (core structure, key data (table), and risk warning) and structured information:

[0102] (1) Core conclusions: Summarize the core trends of BYD's gross profit margin, R&D investment and industry position in the past three quarters;

[0103] (2) Key data: BYD's gross profit margin, R&D investment data and industry comparison are presented in tabular form;

[0104] (3) Risk warning: Point out potential risks that may affect gross profit margin (such as a rebound in raw material prices); the language should be concise and avoid redundancy, and the research report ID should be marked for the data source.

[0105] Understandably, the large language model uses the second prompt word and the research report template to transform the structured research report information into research report content expressed in natural language; then it verifies whether the research report has missed the user's core needs. If there are any missing needs, data is extracted again to supplement and improve the report; if the research report content has fully covered the user's core needs, a personalized research report is finally generated.

[0106] Understandably, if the target users are those who prefer chart presentation, dynamic charts can be embedded in the generation process of personalized research reports, and the client can switch the chart display type when viewing.

[0107] Figure 3 This is a schematic diagram illustrating a personalized research report provided in an embodiment of this application. (Combined with...) Figure 3 It can be seen that personalized research reports may include core conclusions, indicators, target company (such as BYD) gross profit margin, and risk warnings as a second relevant piece of information.

[0108] In one alternative implementation, the research report push period can be determined based on the work rhythm characteristics and the timeliness of the information in the personalized research report; within the research report push period, the personalized research report is pushed to the target user.

[0109] The timeliness of information in personalized research reports is categorized into high-level, medium-level, and standard-level. High-level information corresponds to urgent information, such as sudden policy changes; medium-level information corresponds to highly time-sensitive information, such as corporate financial reports; and standard-level information corresponds to general information, such as monthly industry analysis reports.

[0110] For example, if the timeliness of the information in the personalized research report is high, the target time period starting from the current moment will be used as the time period for pushing the research report; if the timeliness of the information in the personalized research report is medium, the time period for pushing the research report will be determined based on the characteristics of the work rhythm.

[0111] Taking user A as an example, assuming that user A processes research reports from 9 am to 12 pm, then for medium-level information, the research report push time period is determined to be from 9 am to 12 pm.

[0112] It should be noted that between 9:00 AM and 12:00 PM, medium-level information will be sent to user A first; then, ordinary-level research reports will be sent in order of publication time.

[0113] It should be noted that if a research report on the same topic has already been pushed to the target user within 3 consecutive hours, the target user will not be pushed the same research report again to avoid information overload.

[0114] It should be emphasized that if the target user searches for research report data of the target company in real time, the timeliness of the information related to the target company in the personalized research report will be modified to the higher level and pushed to the user with priority.

[0115] In one optional implementation, after the service pushes the aforementioned personalized research report information to the user, it records the report push status, including delivered, read, not delivered, and unread.

[0116] While reading personalized research reports, users can click on the matching degree evaluation (complete match, partial match, and no match) in the client and fill in optimization suggestions, etc.

[0117] This application can further obtain feedback information from target users on personalized research reports, and update user demand profiles and research report templates based on the feedback information.

[0118] In one alternative implementation, the system can be iteratively optimized every two weeks, specifically including the following three adjustments:

[0119] Large model and prompt word optimization: Based on "information extraction error cases" (such as missing enterprise operation indicators and core industry data that users care about), the parameters of the large language model are fine-tuned and the prompt word template is optimized at the same time to improve the extraction accuracy of structured research report information and reduce demand matching deviation.

[0120] Dynamic adjustment of push time: Push strategies are formulated based on the "read rate" data of user research reports. For example, if user A's research report read rate is 80% in the morning and only 50% in the afternoon, the push time of the user's research report will be adjusted to the morning to increase the probability of effective reading after the research report is reached.

[0121] Demand Graph Weight Iteration: Adjust the feature weights of the user demand graph based on the "matching score" between the research report content and user needs. For example, if users give a low matching score to the "R&D investment" related content in the research report, it means that the information coverage of this indicator is insufficient. Then expand the information extraction range of the "R&D investment" dimension to enhance the adaptability of this type of content to user needs.

[0122] After completing the above optimizations, the adjusted large model parameters, the updated user demand map, and the optimized push strategy will be synchronized to the corresponding database to achieve continuous iterative upgrades of system functions.

[0123] It should be noted that before generating personalized research reports in this application, the large language model needs to be deployed on the server's GPU node; research report data sources (brokerage research report platform API), market data interfaces (including corporate financial report data and industry indicator data), and user behavior collection interfaces (including client reading records and labeled data) need to be established, and the addresses and interface parameters of each data source need to be uniformly stored in the system resource library; at the same time, the initialization of the user demand database, research report information database, and personalized research report database needs to be completed, which will be used to store user demand information, extracted structured research report data, and generated personalized research report results, respectively.

[0124] In summary, the personalized research report generation method disclosed in this application has the following advantages:

[0125] First, in the process of building user demand profiles, we break through the limitations of traditional keyword matching, integrate static user preference data and dynamic behavioral data, and construct a structured demand map that covers industry preferences, indicator focus, reading habits, and information acquisition rhythm; at the same time, we combine real-time user feedback for dynamic iteration and optimization to ensure that the research report information screening logic is deeply aligned with the user's real needs.

[0126] Second, based on a pre-trained large language model in the financial field, it can achieve a semantic-level deep understanding of the content of research reports, accurately distinguish the sub-sectors that users are interested in (such as "new energy vehicle companies" and "upstream of new energy") and the level of information importance, and effectively avoid irrelevant information redundancy.

[0127] Third, it supports dynamically adjusting the presentation structure, data display format, and content detail level of research reports based on users' reading habits and content preferences, building a personalized generation mode of "one template per person," reducing the information understanding cost for investment research personnel, and adapting to the work styles and scenario needs of different users.

[0128] Fourth, based on users' work rhythm, real-time needs, and information timeliness, a tiered push strategy should be developed: emergency information should be pushed immediately, routine information should be pushed out of peak hours, and real-time needs should be prioritized.

[0129] Fifth, based on user feedback data, push effect data, and information extraction accuracy data, the large language model, user demand graph, and push strategy are dynamically iteratively optimized to build a self-iterative closed loop for the system, enabling the system to maintain high accuracy and high adaptability in the long term without the need for frequent manual intervention and adjustment.

[0130] Based on the same inventive concept, this application provides a personalized research report generation device. Figure 4 This is a schematic diagram of a personalized research report generation device provided in an embodiment of this application. (Combined with...) Figure 4 As shown, the personalized research report generation device 400 in this application includes:

[0131] The demand profile determination module 401 is used to create a user demand profile; the user demand profile includes reading habit characteristics, content preference characteristics, and analysis dimension preference characteristics.

[0132] The research report text determination module 402 is used to obtain the preprocessed research report text;

[0133] The structured research report information generation module 403 is used to generate structured research report information through a large language model, the preprocessed research report text, the user demand profile, and the first prompt word;

[0134] The research report template determination module 404 is used to determine the research report template based on the reading habit characteristics and the content preference characteristics;

[0135] The research report generation module 405 is used to generate personalized research reports using the large language model, the research report template, the structured research report information, the analysis dimension preference features, and the second prompt words.

[0136] In one alternative implementation, the personalized research report generation device 400 further includes:

[0137] The push period determination module is used to determine the push period of the research report based on the work rhythm characteristics and the timeliness of the information in the personalized research report;

[0138] The research report push module is used to push the personalized research report to the target user during the research report push period.

[0139] In one alternative implementation, the requirement profiling module 401 includes:

[0140] The first data acquisition unit is used to acquire the demand questionnaire data filled out by the target user; the demand questionnaire data includes the industry sub-sectors that the user is interested in, the enterprise operation indicators that the user is interested in, the user's reading habits, and the user's work rhythm.

[0141] The second data acquisition unit is used to collect reading behavior data of the target user when reading the research report;

[0142] The third data acquisition unit is used to extract information search behavior data of the target user within a preset time period;

[0143] The profile generation unit is used to create the user profile based on the demand questionnaire data, the reading behavior data, and the information search behavior data.

[0144] In one alternative implementation, the research report text determination module 402 includes:

[0145] The first research report data acquisition unit is used to acquire first research report data through an API interface at a preset data acquisition frequency.

[0146] The second research report data acquisition unit is used to process the first research report data according to a preset research report layout format to generate the second research report data; the preset research report layout format includes arranging the data in the order of research report number, publication time, industry, company, and full text.

[0147] The research report text generation unit is used to perform text parsing and text annotation on the second research report data to obtain the preprocessed research report text.

[0148] In one alternative implementation, it includes:

[0149] The first time period determination unit is used to determine the target time period starting from the current time as the time period for pushing the research report if the timeliness of the information in the personalized research report is of a high level.

[0150] The second time period determination unit is used to determine the time period for pushing the research report based on the work rhythm characteristics if the timeliness of the personalized research report is at a medium level.

[0151] In one optional implementation, the push time period determination module further includes:

[0152] The level update unit is used to modify the timeliness of the information related to the target company in the personalized research report information to the higher level if the target user searches for research report data of the target company in real time.

[0153] In one alternative implementation, the personalized research report generation device 400 further includes:

[0154] The feedback information acquisition module is used to acquire the feedback information of the target user on the personalized research report;

[0155] The profile template update module is used to update the user demand profile and the research report template based on the feedback information.

[0156] Based on the personalized research report generation method and apparatus provided in the foregoing embodiments, this application also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements some or all of the steps in the personalized research report generation method mentioned above.

[0157] Based on the personalized research report generation method and apparatus provided in the foregoing embodiments, this application also provides an electronic device, including:

[0158] A memory on which computer programs are stored;

[0159] A processor is configured to execute the computer program in the memory to implement some or all of the steps in the personalized research report generation method provided in the foregoing embodiments.

[0160] It should be noted that the various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, for the device embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can be referred to the description of the method embodiments. The device embodiments described above are merely illustrative, and the units described as separate components may or may not be physically separate. The components indicated as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment solution according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0161] The above description is merely one specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for generating personalized research reports, characterized in that, The method includes: Obtain a user demand profile of the target users; the user demand profile includes reading habit characteristics, content preference characteristics, and analysis dimension preference characteristics. Obtain the preprocessed research report text; Structured research report information is generated using a large language model, the preprocessed research report text, the user demand profile, and the first prompt word. Based on the aforementioned reading habit characteristics and content preference characteristics, a research report template is determined; Personalized research reports are generated using the large language model, the research report template, the structured research report information, the analytical dimension preference features, and the second prompt words.

2. The method according to claim 1, characterized in that, The user demand profile also includes work rhythm characteristics, and the method further includes: Based on the work rhythm characteristics and the timeliness of the information in the personalized research report, the time period for pushing the research report is determined; During the specified research report push period, the personalized research report is pushed to the target user.

3. The method according to claim 1, characterized in that, The process of obtaining the user demand profile of the target user includes: Obtain the demand questionnaire data filled out by the target users; the demand questionnaire data includes the industry sub-sectors that users are interested in, the enterprise operation indicators that users are interested in, the users' reading habits, and the users' work rhythm; Collect reading behavior data of the target users when reading research reports; Extract information search behavior data of the target user within a preset time period; Based on the demand questionnaire data, the reading behavior data, and the information search behavior data, the user demand profile is drawn.

4. The method according to claim 1, characterized in that, The process of obtaining the preprocessed research report text includes: Data from the first research report is collected via API interface at a preset data collection frequency. The first research report data is processed according to a preset research report format to generate the second research report data; the preset research report format includes arranging the data in the order of research report number, publication time, industry, company, and full text. The second research report data is parsed and annotated to obtain the preprocessed research report text.

5. The method according to claim 2, characterized in that, The determination of the research report push time period based on the work rhythm characteristics and the timeliness of information in the personalized research report includes: If the timeliness of the information in the personalized research report is high, then the target time period starting from the current time will be used as the time period for pushing the research report. If the timeliness of the personalized research report is at a medium level, the time period for pushing the research report is determined based on the work rhythm characteristics.

6. The method according to claim 5, characterized in that, The method further includes: If the target user searches for research report data of the target company in real time, the timeliness of the information related to the target company in the personalized research report information will be modified to the higher level.

7. The method according to any one of claims 1-6, characterized in that, Also includes: Obtain feedback information from the target user regarding the personalized research report; Based on the feedback information, update the user demand profile and the research report template.

8. A personalized research report generation device, characterized in that, The device includes: The demand profile determination module is used to create user demand profiles; the user demand profiles include reading habit characteristics, content preference characteristics, and analysis dimension preference characteristics. The research report text determination module is used to obtain the preprocessed research report text; The structured research report information generation module is used to generate structured research report information using a large language model, the preprocessed research report text, the user demand profile, and the first prompt word. The research report template determination module is used to determine the research report template based on the reading habit characteristics and the content preference characteristics; The research report generation module is used to generate personalized research reports using the large language model, the research report template, the structured research report information, the analysis dimension preference features, and the second prompt words.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the program implements the steps of the method described in any one of claims 1-7.

10. An electronic device, characterized in that, include: A memory on which computer programs are stored; A processor for executing the computer program in the memory to implement the steps of the method according to any one of claims 1-7.

Images