Information analysis method and device based on large language model

Abstract

The embodiment of the invention provides an information analysis method and device based on a large language model, and the method comprises the steps: obtaining original data collected through a plurality of channels in a historical time period, obtaining target key information according to the original data, and obtaining target information according to the target key information; according to the method, a plurality of pieces of summary information and a plurality of pieces of expected information corresponding to a plurality of themes in a future time period are generated, and an expected time sequence diagram corresponding to each theme is generated according to the expected information corresponding to each theme in the future time period, so that a large amount of text content is analyzed based on a preset large language model; therefore, the difficulty in regular and comprehensive tracking analysis caused by insufficient manual processing capability is overcome, and the objectivity and fairness of content summarization are improved.

Description

Technical Field

[0001] The present invention relates to the field of automatic information processing, and in particular to an information analysis method and device based on a large language model. Background Art

[0002] At present, each chief economic analyst publishes more than 20 reports per month on average, with a report summary of nearly 50,000 words, a large amount of data and complex analysis logic. In the current actual business scenario, a large amount of text content needs to be processed and analyzed for the monthly report summary, and manual processing capabilities may be insufficient. Therefore, the company's macro researchers mainly track the opinions of other sell-side chief analysts in a "point-like" and subjective manner. Researchers usually manually select and read some reports of certain analysts to conduct analysis and summary, which is very prone to information errors and deviations in analysis conclusions, making it difficult to achieve regular and comprehensive tracking and summary of analysts' opinions; and the research process is mainly based on subjective analysis, and there may be subjective biases when evaluating opinions, which affects the objectivity and fairness of the evaluation. Summary of the invention

[0003] In view of the above problems, a method and device for information analysis based on a large language model is proposed to overcome the above problems or at least partially solve the above problems, including:

[0004] An information analysis method based on a large language model, characterized in that the method comprises:

[0005] Acquire raw data collected through multiple channels within a historical period, and obtain target key information based on the raw data;

[0006] Generate a plurality of summary information and a plurality of expected information corresponding to a plurality of topics in a future time period according to the target key information through a preset large language model;

[0007] According to the expected information corresponding to each topic in the future time period, an expected timing diagram corresponding to each topic is generated; wherein the expected timing diagram includes the expected information corresponding to the topic in the future time period and the expected information in multiple historical time periods.

[0008] Optionally, generating a plurality of expected information corresponding to a plurality of topics in a future time period includes:

[0009] Based on the summary information corresponding to each topic in the future time period, determine the expected information corresponding to each topic in the future time period.

[0010] Optionally, obtaining target key information according to the original data includes:

[0011] Performing data cleaning on the original data to obtain multiple target data;

[0012] The multiple target data are spliced ​​together to obtain target key information.

[0013] Optionally, generating a plurality of summary information corresponding to a plurality of topics in a future time period according to the target key information includes:

[0014] Based on multiple topic framework information corresponding to multiple topics, the target key information is processed to generate multiple summary information corresponding to the multiple topics in the future time period; wherein the topic framework information includes topic definition, relevant keywords and multiple relevant dimensions.

[0015] Optionally, generating an expected time sequence diagram corresponding to each topic according to expected information corresponding to each topic in a future time period includes:

[0016] Get prospective information on each topic over multiple historical time periods;

[0017] Quantify the expected information corresponding to each topic in the future time period and the expected information in multiple historical time periods to obtain multiple quantitative values ​​corresponding to each topic;

[0018] According to the multiple quantitative values ​​corresponding to each topic and the corresponding historical time period, an expected time series diagram corresponding to each topic is generated.

[0019] Optionally, after generating the expected time sequence diagram corresponding to each topic according to the expected information corresponding to each topic in the future time period, the method further includes:

[0020] Acquire multiple average expected time series diagrams corresponding to the multiple topics; wherein the average expected time series diagrams include average expected information corresponding to the corresponding topics in the future time period and average expected information in multiple historical time periods;

[0021] Generate an expected comparative time series diagram for each subject based on the expected time series diagram corresponding to each subject and the average expected time series diagram;

[0022] Publish multiple timing diagrams of the expected comparison.

[0023] Optionally, the multiple target data include one or more of the following:

[0024] Date data, title data, summary data.

[0025] An information analysis device based on a large language model, the device comprising:

[0026] A target key information acquisition module is used to acquire original data collected through multiple channels within a historical time period, and obtain target key information based on the original data;

[0027] A summary and expected information acquisition module is used to generate a plurality of summary information corresponding to a plurality of topics in a future time period according to the target key information through a preset large language model, and determine the expected information corresponding to each topic in the future time period according to the summary information corresponding to each topic in the future time period;

[0028] The expected timing diagram generation module is used to generate an expected timing diagram corresponding to each topic according to the expected information corresponding to each topic in the future time period; wherein the expected timing diagram includes the expected information corresponding to the topic in the future time period and the expected information in multiple historical time periods.

[0029] An electronic device comprises a processor, a memory and a computer program stored in the memory and capable of running on the processor, wherein when the computer program is executed by the processor, the information analysis method based on a large language model as described above is implemented.

[0030] A readable storage medium stores a computer program, and when the computer program is executed by a processor, the information analysis method based on a large language model as described above is implemented.

[0031] The embodiments of the present invention have the following advantages:

[0032] In an embodiment of the present invention, raw data collected through multiple channels in a historical time period is obtained, and target key information is obtained based on the raw data. A plurality of summary information and a plurality of expected information corresponding to a plurality of topics in a future time period are generated based on the target key information through a preset large language model. An expected time series diagram corresponding to each topic is generated based on the expected information corresponding to the topic in the future time period. The expected time series diagram includes the expected information corresponding to the topic in the future time period and the expected information in multiple historical time periods. This realizes analysis of a large amount of text content based on a preset large language model, and completes systematic qualitative and quantitative analysis of macro-viewpoints, thereby overcoming the difficulty in regular and comprehensive tracking and analysis caused by insufficient manual processing capabilities, and improving the objectivity and fairness of the analysis of analysts' views. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] In order to more clearly illustrate the technical solution of the present invention, the accompanying drawings required for use in the description of the present invention will be briefly introduced below. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative labor.

[0034] Figure 1is a flowchart of the steps of an information analysis method based on a large language model provided by some embodiments of the present invention;

[0035] Figure 2 is a flowchart of steps of an analysis method of an analyst economic report provided by some embodiments of the present invention;

[0036] Figure 3 It is a structural block diagram of an information analysis device based on a large language model provided by some embodiments of the present invention. DETAILED DESCRIPTION

[0037] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0038] Reference Figure 1 , shows a flowchart of a method for analyzing information based on a large language model provided by some embodiments of the present invention, which may specifically include the following steps:

[0039] Step 101, obtaining original data collected through multiple channels within a historical time period, and obtaining target key information based on the original data.

[0040] The original data collected within a historical time period can be collected through multiple channels, that is, the data that needs to be analyzed and summarized, so that the target key information can be obtained based on the collected original data.

[0041] In practical applications, macroeconomic reports of target economic analysts within a certain historical period can be collected from the Zhiqiu research report database. The historical period can be one month. For example, multiple macroeconomic reports released by the economic analyst within this month can be collected at the end of this month. The collected macroeconomic reports can be used as raw data, and information for analysis and summary, i.e., target key information, can be obtained based on the collected macroeconomic reports. In addition, this method can be applied to a unified investment research platform, so that the platform can automatically collect raw data, i.e., multiple macroeconomic reports released by the target economic analyst this month.

[0042] In some embodiments of the present invention, obtaining target key information according to the original data includes:

[0043] Sub-step 11, performing data cleaning on the original data to obtain multiple target data.

[0044] The collected raw data can be cleaned first to obtain multiple key target data.

[0045] In some embodiments of the present invention, the plurality of target data includes one or more of the following:

[0046] Date data, title data, summary data.

[0047] Specifically, data cleaning can be performed on multiple macroeconomic reports released by the target economic analyst this month, thereby removing unnecessary information and leaving necessary information. That is, from multiple macroeconomic reports released by the target economic analyst this month, the writing date, title information and summary information of each macroeconomic report, namely, date data, title data and summary data, are obtained.

[0048] Sub-step 21, splicing the multiple target data to obtain target key information.

[0049] After acquiring multiple target data, the multiple target data can be spliced ​​together, and the spliced ​​data is the target key information.

[0050] Specifically, the writing dates, title information, and summary information of multiple macroeconomic reports can be spliced ​​together, and the target key information is obtained after the splicing is completed.

[0051] Step 102, using a preset large language model, based on the target key information, generates a plurality of summary information and a plurality of expected information corresponding to a plurality of topics in a future time period.

[0052] After obtaining the target key information, the target key information may be input into a preset large language model, so that a plurality of summary information and a plurality of expected information corresponding to a plurality of topics in a future time period may be generated through the large language model.

[0053] Specifically, after obtaining the spliced ​​information formed by the writing date, title information and summary information of multiple macroeconomic reports, the spliced ​​information can be input into the preset large language model. Through the large language model, the spliced ​​information, that is, the views of the target economic analyst in multiple macroeconomic reports this month, is summarized and analyzed from four themes, so as to obtain the four main viewpoints summary information corresponding to the four themes in the future time period of the target analyst, that is, multiple summary information. Among them, the future time period can be the next month, and the four themes can be inflation, financial credit, economic growth and macroeconomic policy.

[0054] Furthermore, through the large language model, the trends of the four themes can be judged based on the original text or summary information of multiple macroeconomic reports of the target economic analysts this month, so as to obtain the target analysts' corresponding four trend expected information for the four themes in the future time period, that is, multiple expected information.

[0055] In some preferred embodiments of the present invention, after the summary information is generated, multiple expected information may be generated based on the summary information, including:

[0056] Based on the summary information corresponding to each topic in the future time period, determine the expected information corresponding to each topic in the future time period.

[0057] That is, after obtaining multiple summary information corresponding to multiple topics in the future time period, the large language model can also generate corresponding multiple expected information based on the multiple summary information, that is, the expected information corresponding to each topic in the future time period.

[0058] Specifically, after obtaining the four main viewpoint summary information corresponding to the four topics in the next month, the four trend expectation information corresponding to the four main viewpoint summary information can be determined through the preset large language model, that is, the trend expectation information corresponding to each topic in the next month, that is, the expectation information. Among them, the trend expectation information can be one of the three trend expectation conclusions, and the trend expectation conclusions can include upward, neutral and downward.

[0059] In some embodiments of the present invention, generating a plurality of summary information corresponding to a plurality of topics in a future time period according to the target key information includes:

[0060] Based on multiple topic framework information corresponding to multiple topics, the target key information is processed to generate multiple summary information corresponding to the multiple topics in the future time period; wherein the topic framework information includes topic definition, relevant keywords and multiple relevant dimensions.

[0061] In practical applications, the preset big language model in the present invention integrates the self-developed macro research and analysis framework. The macro research and analysis framework combines the big language model, that is, with the help of the big model's powerful natural language processing, reasoning, and generation capabilities, combined with the macro research and analysis framework based on macro researchers, to develop algorithms and continuously train the model to enable it to understand the analysis ideas of macro researchers. The opinions of each economic analyst can be based on the theme framework information corresponding to the four themes, and the splicing information formed by the writing dates, title information, and summary information of multiple macroeconomic reports, that is, the target key information, is analyzed and processed, thereby generating four main viewpoint summary information corresponding to the four themes in the next month, that is, multiple summary information.

[0062] Among them, the subject framework information may include subject definition, relevant keywords and multiple relevant dimensions. The relevant keywords may be keywords related to the corresponding subject, and the multiple relevant dimensions may be dimensions associated under the corresponding subject. For example, the multiple relevant dimensions of inflation may include one or more of cause analysis, impact analysis, measurement indicators, response measures and international comparison; the relevant dimensions of financial credit may include one or more of credit subjects, credit instruments, credit risks, credit ratings, credit systems and credit supervision; the relevant dimensions of economic growth may include one or more of measurement indicators, influencing factors, growth models, economic growth theories, economic growth and sustainable development, and economic growth and income distribution; the relevant dimensions of macroeconomic policies may include one or more of policy objectives, policy instruments, policy transmission mechanisms, policy coordination, policy effectiveness evaluation and policy reforms.

[0063] Step 103, generating an expected timing diagram corresponding to each topic according to the expected information corresponding to each topic in the future time period; wherein the expected timing diagram includes the expected information corresponding to the topic in the future time period and the expected information in multiple historical time periods.

[0064] Specifically, after obtaining the four expected trend information corresponding to the four themes in the next month, the expected trend information corresponding to each theme in the next month can be combined to generate the expected time series diagram corresponding to each theme, so as to obtain four expected time series diagrams. Among them, the expected time series diagram can include the expected trend information of a theme in the next month and the expected trend information of the previous months, that is, the expected time series diagram corresponding to each theme can be a curve chart with the horizontal axis of the coordinate axis being different months and the vertical axis being the expected trend information corresponding to different months.

[0065] In some embodiments of the present invention, generating an expected timing diagram corresponding to each topic according to expected information corresponding to each topic in a future time period includes:

[0066] Sub-step 21, obtaining expected information of each topic in multiple historical time periods.

[0067] In practical applications, the technical solution of the present application can be applied to a unified investment research platform, which can store four trend forecast information corresponding to four themes by target economic analysts in multiple historical time periods, where the multiple historical time periods can be multiple months before this month. Thus, the trend forecast information corresponding to each theme in multiple months before this month can be obtained.

[0068] Sub-step 22, quantifying the expected information corresponding to each theme in the future time period and the expected information in multiple historical time periods to obtain multiple quantitative values ​​corresponding to each theme.

[0069] Specifically, after obtaining the expected trend information corresponding to each theme in the months before this month, the expected trend information corresponding to each theme in the months before this month and the expected trend information corresponding to the next month are quantified to obtain the quantitative value of the expected trend information corresponding to each theme in different months, that is, to obtain multiple quantitative values ​​corresponding to each theme. It should be noted that the expected trend information corresponding to each theme in the months before this month includes the expected trend information of this month, and what is obtained this month is the expected trend information for the next month.

[0070] The trend expectation information may be one of upward, neutral or downward. After quantification, the quantization value corresponding to the upward trend may be 1, the quantization value corresponding to the neutral trend may be 0, and the quantization value corresponding to the downward trend may be -1.

[0071] In this way, through quantitative processing, the expected trend information of each topic in multiple months before this month and the expected information in the next month are converted into quantitative values ​​of 1, 0, and -1.

[0072] Sub-step 23, generating an expected time series diagram corresponding to each theme according to the multiple quantitative values ​​corresponding to each theme and the corresponding historical time period.

[0073] Specifically, after obtaining multiple quantitative values ​​corresponding to each theme, the expected time series diagram corresponding to each theme is generated in combination with the month corresponding to each quantitative value, so that four expected time series diagrams corresponding to four themes can be obtained, that is, multiple expected time series diagrams are obtained.

[0074] In some embodiments of the present invention, after generating the expected timing diagram corresponding to each topic according to the expected information corresponding to each topic in the future time period, the method further includes:

[0075] Sub-step 31, obtaining multiple average expected time series diagrams corresponding to the multiple topics; wherein the average expected time series diagram includes the average expected information corresponding to the corresponding topic in the future time period and the average expected information in multiple historical time periods.

[0076] Specifically, four average expected time series diagrams corresponding to the four themes may be obtained, and the average expected time series diagrams include the average expected information of the corresponding themes in the next month and the average expected information of the previous several months before this month.

[0077] It should be noted that the four expectation time series charts are generated based on multiple macroeconomic reports released by the target economic analysts, that is, the four expectation time series charts are generated based on the four trend expectation information predicted by the target economic analysts personally. In simple terms, the trend expectation information corresponding to the four themes is an analysis and summary of the target economic analysts' personal views, and the average expectation time series chart is based on the market's average trend expectation or the mainstream market view, that is, the average expectation information.

[0078] Sub-step 32, generating an expected comparative time series diagram for each subject according to the expected time series diagram corresponding to each subject and the average expected time series diagram.

[0079] Specifically, after obtaining the expected time series diagram and the average expected time series diagram corresponding to each subject, the curves corresponding to the expected time series diagram and the average expected time series diagram corresponding to each subject can be placed in the same coordinate system, thereby generating the expected comparative time series diagram of each subject. The expected comparative time series diagram includes two curves corresponding to the expected time series diagram and the average expected time series diagram, and the coordinate system is a coordinate system in which the horizontal axis is the month and the vertical axis is the expected information quantization value.

[0080] Sub-step 33, publishing a plurality of the expected comparison timing diagrams.

[0081] After obtaining the four expected comparison time series charts corresponding to the four themes, the four expected comparison time series charts corresponding to the four themes and the four summary information for the next month can be published on the unified investment research platform.

[0082] In some embodiments of the present invention, before obtaining a plurality of average expected timing diagrams corresponding to the plurality of topics, the method further includes:

[0083] Sub-step 41, obtaining manual normative verification results of multiple expected timing diagrams.

[0084] After obtaining the four expected time series diagrams corresponding to the four themes, the four expected time series diagrams will be manually verified to obtain the verification results. The verification may include verifying the expected trend information of each theme for the next month, that is, the expected trend information generated this month.

[0085] It should be noted that each expected trend result is obtained based on the corresponding main point summary information, that is, the summary information, so the above-mentioned normative verification also includes the verification of the summary information of each theme for the next month.

[0086] Sub-step 42, when the verification result is passed, publishing the plurality of expected timing diagrams and a plurality of summary information corresponding to the plurality of topics in the future time period.

[0087] When the verification result is passed, the four expected time series diagrams corresponding to the four themes and the four summary information for the next month can be used as information to be released on the unified investment research platform.

[0088] In one example, when the verification result obtained is verification failure, the large language model integrated with the macro research and analysis framework can be optimized.

[0089] In practical applications, the information analysis method based on the large language model in the present invention can be applied to multiple economic analysts, such as the nine mainstream chief economic analysts in the market, to automatically summarize the views on multiple topics and the corresponding trend expectations every month, and finally convert them into quantitative indicators.

[0090] Furthermore, users of the unified investment research platform can select an economic analyst to view the economic analyst's monthly summary of views and future trend expectations on financial credit, economic growth, inflation, and macroeconomic policies, as well as changes in views this month compared to last month. In addition, users can select a date range to view the economic analyst's trend expectation time series in each dimension and compare it with the market average. They can also compare the similarities and differences between the target economic analyst and other economic analysts or the mainstream market, and conduct historical retrospective analysis to verify the accuracy of the analysis, and further support future view tracking, which greatly expands the depth and breadth of macroeconomic research and realizes research methods that were previously difficult to achieve manually. Moreover, relevant data can be stored in the unified investment research platform, realizing the knowledge accumulation of the unified investment research platform.

[0091] In an embodiment of the present invention, raw data collected through multiple channels in a historical time period is obtained, and target key information is obtained based on the raw data. A plurality of summary information and a plurality of expected information corresponding to a plurality of topics in a future time period are generated based on the target key information through a preset large language model. An expected time series diagram corresponding to each topic is generated based on the expected information corresponding to the topic in the future time period. The expected time series diagram includes the expected information corresponding to the topic in the future time period and the expected information in multiple historical time periods. This realizes analysis of a large amount of text content based on a preset large language model, and completes systematic qualitative and quantitative analysis of macro-viewpoints, thereby overcoming the difficulty in regular and comprehensive tracking and analysis caused by insufficient manual processing capabilities, and improving the objectivity and fairness of the analysis of analysts' views.

[0092] Reference Figure 2 , shows a flowchart of the steps of an analysis method of an analyst economic report provided by some embodiments of the present invention, in order to enable those skilled in the art to better understand the above steps, the following is combined with the attached Figure 2The embodiments of the present invention are exemplarily described, but it should be understood that the embodiments of the present invention are not limited thereto:

[0093] 1. Obtain multiple economic reports (i.e. original data) of the target economic analyst for the current month from the Zhiqiu Research Report Database;

[0094] 2. Screen multiple economic reports, i.e., clean them, and obtain the writing date, report title, and report summary (i.e., multiple target data) of each economic report;

[0095] 3. Splice the writing dates, report titles and report summaries of multiple economic reports to obtain splicing information (i.e. target key information);

[0096] 4. Input the target key information into the preset large language model, which integrates the self-developed macro research framework;

[0097] 5. The large language model combines the macro research framework to process the target key information based on the topic definition, relevant keywords and multiple related dimensions to obtain the summary of the main viewpoints of each topic for the month (i.e. multiple summary information) and the expected future trend of each topic (i.e. multiple expected information);

[0098] 6. Based on the summary of the main viewpoints of each topic in the current month and the expected future trends of each topic output by the big language model, obtain the analyst's full historical macro-topic trend expected time series chart (i.e. expected time series chart);

[0099] 7. Verify the trend forecast time series chart through the Macro Research Institute and obtain the verification result (i.e. the result of manual verification). If the verification passes, execute step 9; if it fails, execute step 8.

[0100] 8. Optimize the large language model that integrates the macro research framework;

[0101] 9. Publish the expected trend time series chart (i.e. publish multiple expected comparison time series charts).

[0102] It should be noted that, for the sake of simplicity, the method embodiments are described as a series of action combinations, but those skilled in the art should be aware that the embodiments of the present invention are not limited by the order of the actions described, because according to the embodiments of the present invention, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all preferred embodiments, and the actions involved are not necessarily required by the embodiments of the present invention.

[0103] Reference Figure 3, shows a schematic diagram of the structure of an information analysis device based on a large language model provided by some embodiments of the present invention, which may specifically include the following modules:

[0104] The target key information acquisition module 301 is used to acquire the original data collected through multiple channels in the historical time period, and obtain the target key information based on the original data;

[0105] A summary and expected information acquisition module 302 is used to generate a plurality of summary information and a plurality of expected information corresponding to a plurality of topics in a future time period according to the target key information through a preset large language model;

[0106] The expected timing diagram generation module 303 is used to generate an expected timing diagram corresponding to each topic according to the expected information corresponding to each topic in the future time period; wherein the expected timing diagram includes the expected information corresponding to the topic in the future time period and the expected information in multiple historical time periods.

[0107] In some embodiments of the present invention, the summary and expected information acquisition module 302 includes:

[0108] The expected information acquisition submodule is used to determine the expected information corresponding to each topic in the future time period according to the summary information corresponding to each topic in the future time period.

[0109] In one embodiment of the present invention, the target key information acquisition module 301 includes:

[0110] A data cleaning submodule, used for cleaning the original data to obtain multiple target data;

[0111] The splicing submodule is used to splice the multiple target data to obtain target key information.

[0112] In one embodiment of the present invention, the summary and expected information acquisition module 302 includes:

[0113] The summary information acquisition submodule is used to process the target key information based on multiple topic framework information corresponding to multiple topics, and generate multiple summary information corresponding to the multiple topics in the future time period; wherein the topic framework information includes topic definition, relevant keywords and multiple relevant dimensions.

[0114] In one embodiment of the present invention, the expected timing diagram generating module 303 includes:

[0115] The historical data acquisition submodule is used to obtain the expected information of each topic in multiple historical time periods;

[0116] The quantification submodule is used to quantify the expected information corresponding to each topic in the future time period and the expected information in multiple historical time periods to obtain multiple quantitative values ​​corresponding to each topic;

[0117] The expected time sequence diagram generation submodule is used to generate the expected time sequence diagram corresponding to each theme according to the multiple quantitative values ​​corresponding to each theme and the corresponding historical time period.

[0118] In one embodiment of the present invention, after generating the expected time sequence diagram corresponding to each topic according to the expected information corresponding to each topic in the future time period, the method further includes:

[0119] An average expected time series diagram acquisition module is used to acquire multiple average expected time series diagrams corresponding to the multiple topics; wherein the average expected time series diagram includes average expected information corresponding to the corresponding topic in the future time period and average expected information in multiple historical time periods;

[0120] An expected comparison time series diagram generation module is used to generate an expected comparison time series diagram for each subject according to the expected time series diagram corresponding to each subject and the average expected time series diagram;

[0121] A publishing module is used to publish a plurality of the expected comparison timing diagrams.

[0122] In an embodiment of the present invention, the plurality of target data includes one or more of the following:

[0123] Date data, title data, summary data.

[0124] In an embodiment of the present invention, raw data collected through multiple channels in a historical time period is obtained, and target key information is obtained based on the raw data. A plurality of summary information and a plurality of expected information corresponding to a plurality of topics in a future time period are generated based on the target key information through a preset large language model. An expected time series diagram corresponding to each topic is generated based on the expected information corresponding to the topic in the future time period. The expected time series diagram includes the expected information corresponding to the topic in the future time period and the expected information in multiple historical time periods. This realizes analysis of a large amount of text content based on a preset large language model, and completes systematic qualitative and quantitative analysis of macro-viewpoints, thereby overcoming the difficulty in regular and comprehensive tracking and analysis caused by insufficient manual processing capabilities, and improving the objectivity and fairness of the analysis of analysts' views.

[0125] Some embodiments of the present invention further provide an electronic device, comprising a processor, a memory, and a computer program stored in the memory and capable of running on the processor, wherein the above method is implemented when the computer program is executed by the processor.

[0126] Some embodiments of the present invention further provide a computer-readable storage medium, on which a computer program is stored, and the computer program implements the above method when executed by a processor.

[0127] Some embodiments of the present invention further provide a computer program product, including a computer program, which implements the above method when executed by a processor.

[0128] As for the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0129] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with the relevant laws, regulations and standards of the relevant countries and regions, and provide corresponding operation entrances for users to choose to authorize or refuse.

[0130] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0131] Those skilled in the art will appreciate that the embodiments of the present invention may be provided as methods, devices, or computer program products. Therefore, the embodiments of the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the embodiments of the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0132] The embodiments of the present invention are described with reference to the flowcharts and / or block diagrams of the methods, terminal devices (systems), and computer program products according to the embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of the processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing terminal device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing terminal device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0133] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing terminal device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce a manufactured product including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0134] These computer program instructions can also be loaded onto a computer or other programmable data processing terminal device so that a series of operating steps are executed on the computer or other programmable terminal device to produce a computer-implemented process, thereby providing instructions for implementing the process in the computer or other programmable terminal device. Figure 1 A process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0135] Although the preferred embodiments of the present invention have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the embodiments of the present invention.

[0136] Finally, it should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or terminal device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or terminal device. In the absence of further restrictions, the elements defined by the sentence "including one..." do not exclude the existence of other identical elements in the process, method, article or terminal device including the above elements.

[0137] The above is a detailed introduction to the information analysis method and device based on a large language model. Specific examples are used in this article to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core idea; at the same time, for those skilled in the art, according to the idea of ​​the present invention, there will be changes in the specific implementation method and application scope. In summary, the content of this specification should not be understood as a limitation on the present invention.

Claims

1. An information analysis method based on a large language model, characterized in that: The method comprises: Acquire raw data collected through multiple channels within a historical period, and obtain target key information based on the raw data; Generate a plurality of summary information and a plurality of expected information corresponding to a plurality of topics in a future time period according to the target key information through a preset large language model; According to the expected information corresponding to each topic in the future time period, an expected timing diagram corresponding to each topic is generated; wherein the expected timing diagram includes the expected information corresponding to the topic in the future time period and the expected information in multiple historical time periods.

2. The method according to claim 1, characterized in that The generating of a plurality of expected information corresponding to the plurality of topics in a future time period includes: Based on the summary information corresponding to each topic in the future time period, determine the expected information corresponding to each topic in the future time period.

3. The method according to claim 1, characterized in that The step of obtaining target key information according to the original data includes: Performing data cleaning on the original data to obtain multiple target data; The multiple target data are spliced ​​together to obtain target key information.

4. The method according to claim 1, characterized in that: The generating of a plurality of summary information corresponding to a plurality of topics in a future time period according to the target key information includes: Based on multiple topic framework information corresponding to multiple topics, the target key information is processed to generate multiple summary information corresponding to the multiple topics in the future time period; wherein the topic framework information includes topic definition, relevant keywords and multiple relevant dimensions.

5. The method according to claim 1, characterized in that: The generating of the expected time sequence diagram corresponding to each topic according to the expected information corresponding to each topic in the future time period includes: Get prospective information on each topic over multiple historical time periods; Quantify the expected information corresponding to each topic in the future time period and the expected information in multiple historical time periods to obtain multiple quantitative values ​​corresponding to each topic; According to the multiple quantitative values ​​corresponding to each topic and the corresponding historical time period, an expected time series diagram corresponding to each topic is generated.

6. The method according to any one of claims 1 to 5, characterized in that: After generating the expected time sequence diagram corresponding to each topic according to the expected information corresponding to each topic in the future time period, the method further includes: Acquire multiple average expected time series diagrams corresponding to the multiple topics; wherein the average expected time series diagrams include average expected information corresponding to the corresponding topics in the future time period and average expected information in multiple historical time periods; Generate an expected comparative time series diagram for each subject based on the expected time series diagram corresponding to each subject and the average expected time series diagram; Publish multiple timing diagrams of the expected comparison.

7. The method according to claim 3, characterized in that The multiple target data include one or more of the following: Date data, title data, summary data.

8. An information analysis device based on a large language model, characterized in that: The device comprises: A target key information acquisition module is used to acquire original data collected through multiple channels within a historical time period, and obtain target key information based on the original data; A summary and expected information acquisition module, used to generate a plurality of summary information corresponding to a plurality of topics in a future time period according to the target key information through a preset large language model, and determine the expected information corresponding to each topic in the future time period according to the summary information corresponding to each topic in the future time period; The expected timing diagram generation module is used to generate an expected timing diagram corresponding to each topic according to the expected information corresponding to each topic in the future time period; wherein the expected timing diagram includes the expected information corresponding to the topic in the future time period and the expected information in multiple historical time periods.

9. An electronic device, characterized in that: The invention comprises a processor, a memory and a computer program stored in the memory and capable of running on the processor, wherein when the computer program is executed by the processor, the information analysis method based on a large language model as described in any one of claims 1 to 7 is implemented.

10. A readable storage medium, characterized in that: The readable storage medium stores a computer program, and when the computer program is executed by a processor, the information analysis method based on a large language model according to any one of claims 1 to 7 is implemented.

Images