Event detail content analysis method and device, equipment and medium

Abstract

The invention provides an event detail content analysis method and device, equipment and a medium, and relates to the technical field of large language models. The method comprises the following steps: inputting a prompt text into a large language model, and analyzing event detail contents according to specified analysis requirements through the large language model to obtain an event level, a first event classification, an emotion tag and an event description tag of the event detail contents; the prompt text comprises original content, a specified analysis requirement, a sample and an output format; acquiring a corresponding target event title for the target event detail content of which the event level is the target level; the real-time content of a target event title is retrieved and an event analysis report is generated, so that a large language model is introduced in an event analysis and arrangement stage, the large language model assists human in judging, analyzing and arranging event detail content, a large amount of manpower and time required by traditional manual annotation, modeling and machine learning are saved, and the efficiency is improved. The problem that the model needs to be maintained at any time is solved.

Description

Technical Field

[0001] Embodiments of the present invention relate to the technical field of large language models, and in particular to a method, device, equipment and medium for analyzing event details content. Background Art

[0002] Currently, event monitoring refers to: integrating Internet information collection technology and information intelligent processing technology to automatically capture, classify and cluster massive amounts of Internet information, detect topics, and focus on topics, so as to meet users' information needs such as network event monitoring and news topic tracking, and form analysis results such as briefings, reports, and charts, so as to help customers fully grasp the ideological trends of the masses, make correct guidance, and provide analysis basis.

[0003] Event monitoring systems generally need to solve the following problems: 1. Event capture: based on the user's areas of concern, realize real-time discovery of events such as text, pictures, and videos on the entire network; 2. Event details: inform the user of the event details; 3. Event analysis: identify event levels, sentiment classifications, event labels, and event classifications; 4. Event reporting: send reports regularly or in real time.

[0004] To achieve the above functions, traditional event monitoring systems need to implement the following technologies: 1. Event collection through crawlers or APIs combined with keywords and traditional search technologies; 2. Crawling and parsing event details through crawler technology; 3. Through traditional big data technologies, such as PaddlePaddle and TensorFlow, the corpus is collected in advance through manual sorting, classification, labeling, and grading, and various event analysis models are established through machine learning to achieve intelligent analysis and sorting of events.

[0005] At present, the following problems exist in the event analysis and event reporting stages: Traditional big data technologies, such as PaddlePaddle and TensorFlow, are generally used to conduct specific analysis of events. Corpuses are collected manually in advance for classification, labeling, and grading. Various event analysis models are established through machine learning to achieve intelligent analysis and organization of events. As a result, a large amount of manpower is required for labeling, modeling, and evaluation. The analysis results depend on the quality of machine learning, and continuous model optimization and result tracking are required for new events, new hot spots, and new fields. Summary of the invention

[0006] Embodiments of the present invention provide an event details content analysis method, apparatus, device, and medium to overcome the above problems or at least partially solve the above problems.

[0007] A first aspect of an embodiment of the present invention provides a method for analyzing event details content, the method comprising:

[0008] Inputting the prompt text into the large language model, analyzing the event details content according to the specified analysis requirements through the large language model, and obtaining the event level, the first event classification, the emotion label and the event description label of the event details content; the prompt text includes: the original content, the specified analysis requirements, the sample and the output format;

[0009] For the target event details whose event level is the target level, obtaining the corresponding target event title;

[0010] Retrieve the real-time content of the target event title and generate an event analysis report.

[0011] Optionally, the method further comprises:

[0012] Clustering the event details using a clustering algorithm to obtain a second event classification of the event details;

[0013] For the target event details content whose event level is the target level, determining whether the second event classification of the target event details content matches the first event classification;

[0014] The obtaining of the corresponding target event title for the target event details having the target event level as the target level comprises:

[0015] When the second event classification matches the first event classification, the corresponding target event title is obtained for the target event details content whose event level is the target level.

[0016] Optionally, the method further comprises:

[0017] Obtaining scores of multiple users on the event analysis report;

[0018] Determine the content analysis capability of event details based on multiple scores;

[0019] Based on the event detail content analysis capability, the prompt text is modified until the event detail content analysis capability meets a preset condition.

[0020] Optionally, the method further comprises:

[0021] The event analysis report is sent in real time by calling a report sending module through the large language model.

[0022] Optionally, retrieving the real-time content of the target event title includes:

[0023] Calling multiple search engines through the large language model to retrieve the real-time content of the target event title;

[0024] The large language model is used to call a summary module to summarize the real-time contents of the multiple target event titles retrieved by the multiple search engines until the preset content arrangement requirements are met.

[0025] Optionally, aggregating the real-time contents of the multiple target event titles retrieved by the multiple search engines includes:

[0026] Based on the user interaction amounts corresponding to the real-time contents of the multiple target event titles, the real-time contents of the multiple target event titles are sorted, and the real-time contents of the target event titles corresponding to the top K user interaction amounts are summarized;

[0027] The user interaction volume includes at least one of the following: user click volume, user search volume; K is a natural number greater than or equal to 1.

[0028] A second aspect of an embodiment of the present invention provides an event details content analysis device, the device comprising:

[0029] A content analysis module is used to input the prompt text into a large language model, analyze the event details content according to the specified analysis requirements through the large language model, and obtain the event level, first event classification, emotion label and event description label of the event details content; the prompt text includes: original content, the specified analysis requirements, samples and output format;

[0030] A title acquisition module, used for acquiring the corresponding target event title for the target event details whose event level is the target level;

[0031] The report generation module is used to retrieve the real-time content of the target event title and generate an event analysis report.

[0032] Optionally, the device further comprises:

[0033] A clustering module, used for clustering the event details content using a clustering algorithm to obtain a second event classification of the event details content;

[0034] A matching module, for determining, for the target event details whose event level is the target level, whether the second event classification of the target event details matches the first event classification;

[0035] The title acquisition module includes:

[0036] The title acquisition submodule is used to acquire the corresponding target event title for the target event details content whose event level is the target level when the second event classification matches the first event classification.

[0037] Optionally, the device further comprises:

[0038] A scoring determination module, used to obtain scores of multiple users on the event analysis report;

[0039] A capability determination module, for determining event detail content analysis capabilities based on multiple scores;

[0040] The prompt revision module is used to modify the prompt text based on the event detail content analysis capability until the event detail content analysis capability meets a preset condition.

[0041] Optionally, the device further comprises:

[0042] The report sending module is used to call the report sending module through the large language model to send the event analysis report in real time.

[0043] Optionally, the report generating module includes:

[0044] A search module, used to call multiple search engines through the large language model to retrieve the real-time content of the target event title;

[0045] The summarization module is used to call the summarization module through the large language model to summarize the real-time content of the multiple target event titles retrieved by the multiple search engines until the preset content arrangement requirements are met.

[0046] Optionally, the aggregation module includes:

[0047] A content sorting module, used to sort the real-time contents of the multiple target event titles based on the user interaction amounts corresponding to the real-time contents of the multiple target event titles, and summarize the real-time contents of the target event titles corresponding to the first K user interaction amounts;

[0048] The user interaction volume includes at least one of the following: user click volume, user search volume; K is a natural number greater than or equal to 1.

[0049] A third aspect of an embodiment of the present invention provides an electronic device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the computer program is executed by the processor, the event detail content analysis method of the first aspect of the embodiment of the present invention is implemented.

[0050] A fourth aspect of an embodiment of the present invention provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the event detail content analysis method of the first aspect of the embodiment of the present invention is implemented.

[0051] In the event detail content analysis method provided by the embodiment of the present invention, first, the prompt text is input into the large language model, and the event detail content is analyzed by the large language model according to the specified analysis requirements to obtain the event level, the first event classification, the emotion label and the event description label of the event detail content; the prompt text includes: the original content, the specified analysis requirements, the sample and the output format; then, for the target event detail content with the event level being the target level, the corresponding target event title is obtained; finally, the real-time content of the target event title is retrieved and an event analysis report is generated.

[0052] This embodiment introduces a large language model in the event analysis and collation stage of the event monitoring system. The large language model with human language understanding ability assists humans in judging, analyzing and collating event details. It only needs to pre-specify analysis requirements, descriptive words and some samples, saving a lot of manpower and time required for traditional manual labeling, modeling, and machine learning, solving the problem of needing to maintain the model at any time, improving the accuracy of data collation, event collection and recording, reducing the workload of manual participation, and greatly improving work quality and work efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments 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.

[0054] Figure 1 is a flow chart of an event details content analysis method shown in one embodiment of the present invention;

[0055] Figure 2 is a dynamic clustering flow chart provided by an embodiment of the present invention;

[0056] Figure 3 It is an overall flow chart of a method for improving the automatic processing capability of event monitoring based on a large language model according to an embodiment of the present invention;

[0057] Figure 4 is a structural block diagram of an event details content analysis device provided by an embodiment of the present invention;

[0058] Figure 5 It is a schematic diagram of an electronic device shown in one embodiment of the present invention. DETAILED DESCRIPTION

[0059] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of 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.

[0060] Hereinafter, specific examples of the present solution will be described in more detail with reference to the accompanying drawings.

[0061] refer to Figure 1 , Figure 1 FIG. 1 is a flow chart of a method for analyzing event details content according to an embodiment of the present invention. Figure 1 As shown, the event details content analysis method of this embodiment may include the following steps:

[0062] Step S11: Input the prompt text into the large language model, and analyze the event details content according to the specified analysis requirements through the large language model to obtain the event level, first event classification, emotion label and event description label of the event details content; the prompt text includes: original content, the specified analysis requirements, samples and output format.

[0063] In this embodiment, for event analysis, a large language model can be integrated in the event analysis stage. The large language model (LLM) specifically refers to the generative AI (Generative Artificial Intelligence) large model in the document. It is a type of artificial intelligence technology that can generate content, solutions or new concepts. It is usually based on advanced algorithms such as deep learning and neural networks, and through training with large amounts of data, it enables computers to think and create like humans. LLM is different from traditional AI models. Generally, no training is required. The AI ​​capabilities in application projects can be realized through development methods such as prompt word engineering, knowledge base, and agent.

[0064] For event analysis, the collected event details and prompt text can be input into the large language model, and the event details can be analyzed by the large language model according to the specified analysis requirements to obtain the event level, first event classification, emotion label and event description label of the event details.

[0065] The prompt text is used to guide the large language model to output the analysis results corresponding to the event details obtained by analyzing the event details (such as event level, first event classification, emotion label, and event description label, etc.). The prompt text includes at least: original content, specified analysis requirements, samples, and output format. The original content is the event details content input to the large language model, and the first event classification is the event classification obtained by the large language model based on the analysis of the event details content.

[0066] In an optional example, a prompt text, i.e., a series of prompt words and examples, is set for the large language model to implement content analysis of the event level, the first event classification, the emotion label, and the event description label of the event details:

[0067] 1) Event level. First, the event ranking and number of comments obtained from the event source through crawler technology are used as the main basis. Secondly, a large language model is used to provide prompt text for rating, such as: "The following 10 items are divided into four levels, ABCD, according to the degree and scope of impact. Please grade the events according to the examples. Example 1: XXX event, affected XXX, the impact scope is XXX, level A, Example 2: YYY event, affected YYY, the impact scope is YYY, level B."

[0068] 2) The first event classification is based on the event classification obtained from the event source through crawler technology. Secondly, the large language model is used to provide prompt samples for classification, such as: "Please classify the following events. The categories include: product problems, service problems, user complaints, cross-level complaints, asset damage problems, and promotional activities. The examples are as follows: "

[0069] 3) Event description tags: send the latest events (i.e. the latest collected event details) in batches to the large language model, and use the large language model to extract events, such as "Please extract the latest event description tags based on the following events". Combined with historical event tags, the event description tags are annotated through LLM.

[0070] 4) Emotion labeling: Through the large language model, prompt text is provided for automatic labeling, such as: "Please classify the following events according to emotion. The categories include: positive events, negative events, and neutral events. The examples are as follows: "

[0071] In addition, in an optional embodiment, after completing the automatic analysis of events through a large language model, detailed event data (such as event details and their corresponding event levels, first event classifications, emotion labels, and event description labels) are stored in a MySQL database and a vector database for subsequent data display, data query, view display, and event Q&A.

[0072] Step S12: For the target event details whose event level is the target level, obtain the corresponding target event title.

[0073] In this embodiment, after obtaining the event level corresponding to the event details content, it can be determined whether the event level is the target level, and the event details content with the target level is determined as the target event details content, and the corresponding target event title is obtained for the target event details content.

[0074] In this embodiment, a target level is set in advance, and the target level is the level corresponding to the event details of the event analysis report to be generated. For example, the target level is level A, or the target level is level 3, etc. This embodiment does not impose any restrictions on the setting of the target level.

[0075] Step S13: Retrieve the real-time content of the target event title and generate an event analysis report.

[0076] In this embodiment, after obtaining the target event title, the target event title can be searched, more real-time content corresponding to the target event title can be obtained through the Internet, and an event analysis report corresponding to the target event details can be generated to complete the event analysis.

[0077] In this embodiment, a large language model is introduced in the event analysis and collation stage of the event monitoring system. The large language model with human language understanding ability assists humans in judging, analyzing and collating event details. Through the native intelligent capabilities of LLM, data collation and automatic labeling of event levels, event classifications, emotion labels, event description labels, etc. are performed. Only analysis requirements, descriptive words and some samples need to be specified in advance, which saves a lot of manpower and time required for traditional manual labeling, modeling, and machine learning, improves the accuracy of data collation, event collection and recording, reduces the workload of manual participation, and greatly improves work quality and work efficiency.

[0078] In addition, in an optional implementation, the large language model can still be introduced in the event analysis report generation stage. When generating the event analysis report, the target event title and the real-time content of the target event title retrieved through the Internet are handed over to the large language model. The large language model performs further language processing on the real-time content corresponding to the event according to the report prompt text (including at least: preset content organization description words, implementation content, report samples and report output format), such as "Please organize a detailed event briefing of about 2,000 words and related reports from various professional media based on the event title and the following retrieved content" to generate an event analysis report.

[0079] In combination with the above embodiments, in one implementation, an embodiment of the present invention further provides a method for analyzing event details content. Specifically, in addition to the above steps, the method may also include step S21 and step S22, and the above step S12 may specifically include step S23:

[0080] Step S21: clustering the event details using a clustering algorithm to obtain a second event classification of the event details.

[0081] In this embodiment, a clustering algorithm can also be used to cluster the event details to obtain a second event classification of the event details. Among them, the second event classification is an event classification obtained using a clustering algorithm for the event details. In addition, the clustering algorithm used in this embodiment can be a K-means++ algorithm. The K-means++ algorithm is an algorithm for selecting initial values ​​(or "seeds") for the K-means clustering algorithm. The location selection of the K initialized centroids has a great influence on the final clustering results and running time, so it is necessary to select K suitable centroids. The basic idea of ​​the K-means algorithm is still to minimize the K-means sum of squares of deviations, but in the specific implementation, a method of successive iterative correction is adopted.

[0082] Step S22: for the target event details content whose event level is the target level, determining whether the second event classification of the target event details content matches the first event classification.

[0083] In this embodiment, for the target event details content whose event level is the target level, the second event classification corresponding to the target event details content can be determined, and then it can be determined whether the second event classification of the target event details content matches the first event classification of the target event details content. In an optional example, when the first event classification is the same or similar to the second event classification (such as belonging to the same category), it can be determined that the second event classification matches the first event classification.

[0084] Step S23: When the second event classification matches the first event classification, the corresponding target event title is obtained for the target event details content whose event level is the target level.

[0085] In this embodiment, when it is determined that the second event classification of the target event details matches the first event classification of the target event details, the corresponding target event title can be obtained for the target event details with the event level being the target level.

[0086] In an optional implementation, in the data preprocessing stage of event analysis, the K-means++ algorithm is used to perform data aggregation on event levels, event classifications, sentiment labels, and event description label features, and data that may belong to multiple sets are uniquely classified to form a main data set that includes the above sub-data sets to ensure the effectiveness of data training. This is because after data aggregation, the data can be used in the process of event service data management capability evaluation and other possible combined model training processes to determine a unique data set for data that may belong to multiple sub-data sets, thereby ensuring the effectiveness of data training.

[0087] Among them, the K-means++ algorithm can be divided into the following steps:

[0088] Step 1: Randomly select an initial center point u between the data points i ;

[0089] Step 2: For each data point x that has not yet been selected, calculate That is, the distance between the selected point z and the closest center that has already been selected;

[0090] Specifically, first, the dynamic screening steps of several data features such as event level, event classification, emotion label, and event description label are used as K center points. Secondly, the priority of each attribute of event analysis service quality is calculated, the correlation analysis of evaluation indicators is performed, and N measurement indicators with relatively low correlation under each quality attribute are retained, and the comprehensive evaluation indicators of event analysis service quality are retained.

[0091] Step 3: Use the weighted probability distribution to randomly select a new data point as the new center, where the probability of the selected point z is the same as is proportional to;

[0092] Step 4: Repeat steps 2 and 3 until K centers are selected (i.e., j = K);

[0093] Step 5: Now that the initial centers have been selected, continue using standard K-means clustering.

[0094] It can be seen that the center obtained in this way can minimize the sum of squared deviations under the classification conditions at the time of iteration. In step 4, if If this holds true for all j = 1, 2, ..., K, that is, the classification and center position of any point have not changed after iterative correction, then the K-means++ algorithm is said to have converged and can be terminated. Otherwise, go to step 2 to continue correcting the sample and dividing the positions of K center points.

[0095] like Figure 2 As shown, Figure 2 This is a dynamic clustering flow chart provided by an embodiment of the present invention. Figure 2In the process, the initial center point is first selected, and then the initial classification is performed based on the initial center point. It is determined whether the classification is reasonable for the initial classification. If the classification is reasonable, the clustering is terminated. If the classification is unreasonable, the initial classification is modified until the classification is reasonable and the clustering is terminated.

[0096] In combination with the above embodiments, in another implementation, the present invention also provides a method for analyzing event details content. Specifically, in addition to the above steps, the method may also include steps S31 to S33:

[0097] Step S31: Obtain scores of multiple users on the event analysis report.

[0098] In this embodiment, after generating an event analysis report for the target event details, multiple users' ratings of the event analysis report can be obtained. The user's rating of the event analysis report depends on the user's subjective feelings after calling the report service, which can be defined as the average value of the user's evaluation of the event service data management capability, which can be recorded as Q DM Among them, subjective feelings can be determined based on feedback information, which includes evaluation information, consultation, scoring, complaints and suggestions of customer feedback report services.

[0099] Step S32: Determine the event detail content analysis capability based on the multiple scores.

[0100] In this embodiment, the event details content analysis capability for the target event details content can be determined based on multiple scores of the event analysis report by multiple users. Among them, the user's score for the event analysis report can be the user's evaluation result, and the evaluation result can be an element in the set E = {strong, relatively strong, general, relatively weak, weak}, and quantified as E = {5, 4, 3, 2, 1}, then the event details content analysis capability can be calculated according to the following formula:

[0101]

[0102] Among them, DM j Represents the quantitative evaluation value of the j-th user on the event analysis report.

[0103] Step S33: Based on the event detail content analysis capability, modify the prompt text until the event detail content analysis capability meets a preset condition.

[0104] In this embodiment, after obtaining the event detail content analysis capability, the prompt text may be modified based on the event detail content analysis capability until the event detail content analysis capability output by the large language model meets a preset condition.

[0105] In combination with the above embodiments, in one implementation, the present invention also provides a method for analyzing event details content. Specifically, in addition to the above steps, the method may also include step S41:

[0106] Step S41: calling a report sending module through the large language model to send the event analysis report in real time.

[0107] In this embodiment, an agent can be set up, and the large language model calls the report sending module to send the event analysis report in real time, so as to realize the real-time sending of the event analysis report in the event analysis stage. Among them, the report sending module of this embodiment can be implemented by the chain in langchain. Among them, langchain is an AI application development framework, which can give AI applications the ability to load local documents, perform semantic matching on local documents, connect to local databases, connect to local API interfaces, and chain calls to large language models and local capabilities (chain) while connecting to LLM capabilities, and achieve the ability to achieve the operation goals specified by the application through multiple rounds of calls.

[0108] In this embodiment, events are reported and sent through daily reports, real-time notifications, etc., and are defined as the average value of users' evaluation of the event service data management capability based on their subjective feelings after invoking the service, thereby further improving the real-time and accuracy of event analysis.

[0109] In combination with the above embodiments, in one implementation, the present invention also provides an event details content analysis. Specifically, in the method, the "retrieval of the real-time content of the target event title" in the above step S13 may specifically include steps S51 and S52:

[0110] Step S51: calling multiple search engines through the large language model to retrieve the real-time content of the target event title.

[0111] In this embodiment, since the content formats of various event sources are not uniform, it is necessary to retrieve the standardized and easy-to-read detailed content of this event by searching the Internet content. First, through langchain, load the search engine chain (such as baiduchain or bingchain), and then search the real-time content of the event on the Internet through the target event title. For example, the large language model and langchain can be used to search the network content of other content sources on the Internet for the target event title. Specifically, multiple search engines can be called through the large language model to retrieve the real-time content of the target event title.

[0112] Step S52: calling a summary module through the large language model to summarize the real-time contents of the multiple target event titles retrieved by the multiple search engines until the preset content arrangement requirements are met.

[0113] In this embodiment, after obtaining the real-time contents of multiple target event titles retrieved by multiple search engines, the aggregation module can be called through the large language model to aggregate the real-time contents of the multiple target event titles until the preset content organization requirements are met.

[0114] For example, an agent may be set up to implement chain execution of multiple chains, and multiple search engine chains may be used at the same time. The number of contents to be searched by each search engine chain may be set, and these contents may be executed in sequence. Then, a summarization chain (i.e., a summarization module) may be added to summarize the real-time contents of multiple target event titles of multiple search engines until the preset content organization requirements are met, thereby obtaining the real-time contents corresponding to the target event titles for generating event analysis reports.

[0115] In combination with the above embodiments, in one implementation, an embodiment of the present invention further provides a method for analyzing event details content. Specifically, in the method, the "summarizing the real-time content of the multiple target event titles retrieved by the multiple search engines" in the above step S52 may specifically include step S61:

[0116] Step S61: sorting the real-time contents of the multiple target event titles based on the user interaction amounts corresponding to the real-time contents of the multiple target event titles, and summarizing the real-time contents of the target event titles corresponding to the top K user interaction amounts.

[0117] In this embodiment, in order to obtain higher-quality detailed content, the user interaction volume corresponding to the real-time content of multiple target event titles retrieved by multiple search engines can be obtained, and then based on the user interaction volume corresponding to the real-time content of the multiple target event titles, for example, in order from large to small, the real-time content of the multiple target event titles are sorted, and the real-time content of the target event titles corresponding to the first K (K is a natural number greater than or equal to 1) user interaction volumes are summarized as real-time content that requires further content organization.

[0118] The user interaction volume in this embodiment includes at least one of the following: user click volume and user search volume.

[0119] In this embodiment, a large language model is integrated in the event reporting stage, and the large language model tracks the key events throughout their life cycle through Internet search, interface call, message sending, etc., instead of simply sending the event analysis report to the person in charge. Instead, in addition to the traditional event analysis report, the large language model is combined with the interface call, knowledge search, chain call and other capabilities of the agent to analyze and follow up the key events (i.e., the target event details corresponding to the target level) throughout their life cycle, thereby improving the efficiency and effectiveness of event processing.

[0120] In one embodiment, if Figure 3 As shown, Figure 3 This is an overall flow chart of a method for improving the automatic processing capability of event monitoring based on a large language model, which is shown in an embodiment of the present invention. This embodiment aims to improve the accuracy of event monitoring collection, organization, analysis, etc. through the application of LLM technology. Figure 3 In the event analysis phase, first, through event data collection, the langchain Al application framework is called to organize the event details, and then the event details are analyzed through the open source AI big model, and the database / memory database and the vector database. The classification, grading, labeling prompt words and samples are specified in the AI ​​big model (i.e., the big language model), and the event level, event classification, emotion label, event description label, etc. are sorted and automatically labeled to obtain the event level, first event classification, emotion label and event description label output by the big language model. At the same time, the original classification, label and AI automatic classification and labeling are also collected in combination with the event source for verification. Then in the event reporting phase, langchain is called through the AI ​​big model, and langchain is used to search the event details, generate an event analysis report, and the chain is called through the big language model for real-time notification of the report.

[0121] Among them, in an optional example, the event analysis technology stack is: development language: python; LLM: chatglm2, chatglm3, llama2; specify prompt words: specify descriptive words and examples required for event level, event classification, emotion label, and event description label content analysis; database: mysql; vector database engine: FAISS.

[0122] The event reporting technology stack is: development language: python; LLM: chatglm2, chatglm3, llama2; AI application development framework: langchain; crawler and web content parsing framework: Scrapy, Playwright; report sending channel: email, SMS, WeChat interface.

[0123] In this embodiment, by introducing a large language model in each stage of event analysis and reporting of the event monitoring system, a pre-trained large model with human language understanding ability assists humans in judging, analyzing, and organizing events, improves the accuracy of event collection and recording, reduces the workload of manual participation, and greatly improves work quality and efficiency.

[0124] 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.

[0125] Based on the same inventive concept, an embodiment of the present invention provides an event details content analysis device 400. Figure 4 , Figure 4 FIG. 1 is a structural block diagram of an event details content analysis device provided by an embodiment of the present invention. Figure 4 As shown, the device 400 includes:

[0126] The content analysis module 401 is used to input the prompt text into the large language model, analyze the event details content according to the specified analysis requirements through the large language model, and obtain the event level, first event classification, emotion label and event description label of the event details content; the prompt text includes: original content, the specified analysis requirements, samples and output format;

[0127] The title acquisition module 402 is used to acquire the corresponding target event title for the target event details whose event level is the target level;

[0128] The report generation module 403 is used to retrieve the real-time content of the target event title and generate an event analysis report.

[0129] Optionally, the device 400 further includes:

[0130] A clustering module, used for clustering the event details content using a clustering algorithm to obtain a second event classification of the event details content;

[0131] A matching module, for determining, for the target event details whose event level is the target level, whether the second event classification of the target event details matches the first event classification;

[0132] The title acquisition module 402 includes:

[0133] The title acquisition submodule is used to acquire the corresponding target event title for the target event details content whose event level is the target level when the second event classification matches the first event classification.

[0134] Optionally, the device 400 further includes:

[0135] A scoring determination module, used to obtain scores of multiple users on the event analysis report;

[0136] A capability determination module, for determining event detail content analysis capabilities based on multiple scores;

[0137] The prompt revision module is used to modify the prompt text based on the event detail content analysis capability until the event detail content analysis capability meets a preset condition.

[0138] Optionally, the device 400 further includes:

[0139] The report sending module is used to call the report sending module through the large language model to send the event analysis report in real time.

[0140] Optionally, the report generating module includes:

[0141] A search module, used to call multiple search engines through the large language model to retrieve the real-time content of the target event title;

[0142] The summarization module is used to call the summarization module through the large language model to summarize the real-time content of the multiple target event titles retrieved by the multiple search engines until the preset content arrangement requirements are met.

[0143] Optionally, the aggregation module includes:

[0144] A content sorting module, used to sort the real-time contents of the multiple target event titles based on the user interaction amounts corresponding to the real-time contents of the multiple target event titles, and summarize the real-time contents of the target event titles corresponding to the first K user interaction amounts;

[0145] The user interaction volume includes at least one of the following: user click volume, user search volume; K is a natural number greater than or equal to 1.

[0146] Based on the same inventive concept, another embodiment of the present invention provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps in the event detail content analysis method described in any of the above embodiments of the present invention are implemented.

[0147] Based on the same inventive concept, another embodiment of the present invention provides an electronic device 500, such as Figure 5 shown. Figure 5 1 is a schematic diagram of an electronic device according to an embodiment of the present invention. The electronic device includes a memory 502, a processor 501, and a computer program stored in the memory and executable on the processor, and when executed by the processor, the steps in the event detail content analysis method according to any of the above embodiments of the present invention are implemented.

[0148] 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.

[0149] 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.

[0150] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, devices, or computer program products. Therefore, embodiments of the present invention may take the form of complete hardware embodiments, complete software embodiments, or embodiments combining software and hardware. Furthermore, embodiments of the present invention may take the form of computer program products 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.

[0151] 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.

[0152] 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.

[0153] 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 executing on the computer or other programmable terminal device to implement the process. Figure 1 A process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0154] 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.

[0155] 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 "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or terminal device including the elements.

[0156] The above is a detailed introduction to the event details content analysis method, device, equipment and medium provided by the present invention. 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 general technical personnel in this field, 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. A method for analyzing event details content, characterized in that: The method comprises: Inputting the prompt text into the large language model, analyzing the event details content according to the specified analysis requirements through the large language model, and obtaining the event level, the first event classification, the emotion label and the event description label of the event details content; the prompt text includes: the original content, the specified analysis requirements, the sample and the output format; For the target event details whose event level is the target level, obtaining the corresponding target event title; Retrieve the real-time content of the target event title and generate an event analysis report.

2. The method according to claim 1, characterized in that: The method further comprises: Clustering the event details using a clustering algorithm to obtain a second event classification of the event details; For the target event details content whose event level is the target level, determining whether the second event classification of the target event details content matches the first event classification; The obtaining of the corresponding target event title for the target event details having the target event level as the target level comprises: When the second event classification matches the first event classification, the corresponding target event title is obtained for the target event details content whose event level is the target level.

3. The method according to claim 1, characterized in that The method further comprises: Obtaining scores of multiple users on the event analysis report; Determine the content analysis capability of event details based on multiple scores; Based on the event detail content analysis capability, the prompt text is modified until the event detail content analysis capability meets a preset condition.

4. The method according to claim 1, characterized in that: The method further comprises: The event analysis report is sent in real time by calling a report sending module through the large language model.

5. The method according to any one of claims 1 to 4, characterized in that: The retrieving the real-time content of the target event title comprises: Calling multiple search engines through the large language model to retrieve the real-time content of the target event title; The large language model is used to call a summary module to summarize the real-time contents of the multiple target event titles retrieved by the multiple search engines until the preset content arrangement requirements are met.

6. The method according to claim 5, characterized in that The step of aggregating the real-time contents of the multiple target event titles retrieved by the multiple search engines includes: Based on the user interaction amounts corresponding to the real-time contents of the multiple target event titles, the real-time contents of the multiple target event titles are sorted, and the real-time contents of the target event titles corresponding to the top K user interaction amounts are summarized; The user interaction volume includes at least one of the following: user click volume, user search volume; K is a natural number greater than or equal to 1.

7. An event details content analysis device, characterized in that: The device comprises: A content analysis module is used to input the prompt text into a large language model, analyze the event details content according to the specified analysis requirements through the large language model, and obtain the event level, first event classification, emotion label and event description label of the event details content; the prompt text includes: original content, the specified analysis requirements, samples and output format; A title acquisition module, used for acquiring the corresponding target event title for the target event details whose event level is the target level; The report generation module is used to retrieve the real-time content of the target event title and generate an event analysis report.

8. The device according to claim 7, characterized in that The device also includes: A clustering module, configured to cluster the event details using a clustering algorithm to obtain a second event classification of the event details; A matching module, for determining, for the target event details whose event level is the target level, whether the second event classification of the target event details matches the first event classification; The title acquisition module includes: The title acquisition submodule is used to acquire the corresponding event title for the target event details content whose event level is the target level when the second event classification matches the first event classification.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the computer program is executed by the processor, the event details content analysis method according to any one of claims 1 to 6 is implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the event details content analysis method as described in any one of claims 1 to 6 is implemented.

Images