Method, system and equipment for automatically interpreting public opinion event and generating report

Abstract

The invention relates to a public opinion event automatic interpretation and report generation method, system and device, the method is applied to an artificial intelligence system, and the method comprises the following steps: data acquisition; preprocessing the data; storing data; key indexes including a public opinion popularity index, an emotion index, a public opinion participation index, an account propagation index and a hot word index are defined; the component definition comprises a public opinion overview component, a topic sentiment analysis component, a channel analysis component, a propagation account analysis component and a word cloud analysis component, and further comprises an overall trend analysis component and a public opinion suggestion and risk assessment component; component starting logic: setting an initial threshold value for each of other key indexes except the emotion index, starting a corresponding component when the corresponding key index exceeds the threshold value after calculation, and starting a topic emotion analysis component, an overall trend analysis component and a public opinion suggestion and risk assessment component when the public opinion summary component is started; component analysis; generating a report; and dynamically adjusting.

Description

Technical Field

[0001] The present invention relates to the technical field of AI intelligent systems, and in particular to a method, system and device for automatically interpreting and reporting public opinion events. Background Art

[0002] Among the existing public opinion report generation technologies, intelligent report generation systems are gradually being applied to public opinion analysis, but there are still significant deficiencies. Current technologies mainly rely on static templates and manual intervention, and cannot adapt to the dynamically changing public opinion environment in real time.

[0003] Among them, the existing technology usually relies on preset report templates to generate public opinion reports by filling the data analysis results into the template. This method includes several stages of data collection, analysis and report generation. The system collects data from social media, news websites, etc. through crawler technology and API interfaces, and then performs topic identification and trend analysis to finally generate a report in a fixed format. Although this method ensures the structure of the report, it has significant deficiencies in flexibility and personalization. Moreover, the existing technology mainly relies on traditional natural language processing (NLP) methods in data processing, including word segmentation, topic extraction and trend identification. These methods are effective in dealing with relatively simple public opinions, but when faced with complex data, they often fail to capture deeper trend changes. In addition, the current report generation system lacks automated intelligent adjustment capabilities. When public opinion changes rapidly, the existing technology cannot automatically adjust the report content according to the new data, and manual intervention is still required for modification and updating, resulting in delayed report generation.

[0004] Therefore, although the existing technology has made some progress in generating public opinion reports, it still has obvious deficiencies in flexibility, real-time and intelligence, which are mainly reflected in the following aspects:

[0005] 1. Lack of flexibility: the existing report generation system relies on fixed templates and lacks personalized customization capabilities. This makes it difficult for the generated reports to meet the specific needs of different users. Users cannot flexibly select the required analysis content and presentation methods, which affects the practicality of the report.

[0006] 2. Poor real-time performance. Existing technologies usually generate reports after collecting and analyzing data, which results in a long generation cycle and is unable to reflect the dynamic changes of public opinion in real time. In the case of rapid evolution of public opinion, users expect to obtain relevant information immediately when an event occurs, and existing systems often miss the best monitoring opportunity.

[0007] 3. Lack of automated adjustment capabilities. The current system cannot dynamically adjust the report content and structure based on real-time data, and relies on manual modification and updating, resulting in delayed analysis results. The system lacks an automated feedback mechanism and cannot quickly adapt to changes in public opinion. Summary of the invention

[0008] In order to solve the problems existing in the above-mentioned prior art, one of the purposes of the present invention is to provide a method for automatically interpreting and reporting public opinion events, which defines a series of intelligent analysis components to flexibly respond to changes in public opinion, and establishes a dynamic feedback mechanism to enhance the adaptive ability of the system. The second purpose of the present invention is to provide a system for automatically interpreting and reporting public opinion events. The system dynamically selects corresponding components according to key indicators through comprehensive activation logic to ensure that report generation can comprehensively cover all aspects of public opinion, and can automatically adjust the activation threshold of each component according to real-time public opinion changes, thereby improving the ability to respond to emergencies. The third purpose of the present invention is to provide a device for automatically interpreting and reporting public opinion events, so that the system can significantly improve the flexibility and real-time performance of public opinion analysis through the combination of intelligent components and dynamic feedback mechanisms.

[0009] The method for automatically interpreting and reporting public opinion events described in the present invention is applied to an artificial intelligence system and includes:

[0010] Data collection: real-time collection of data from news websites, social media and specific industries;

[0011] Data preprocessing: Convert data in different formats into a standard format, remove invalid, duplicate and noisy data, and extract text data from images;

[0012] Data storage: load the pre-trained model, convert text data into high-dimensional semantic vectors and store them;

[0013] Definition of key indicators, including public opinion heat index, sentiment index, public opinion participation index, account communication index and hot word index;

[0014] Component definitions include the public opinion overview component, topic sentiment analysis component, channel analysis component, communication account analysis component, and word cloud analysis component, which correspond to each key indicator. They also include the overall trend analysis component and the public opinion suggestion and risk assessment component.

[0015] Component activation logic: set initial thresholds for each key indicator other than the sentiment indicator. When the corresponding key indicator exceeds the threshold after calculation, the corresponding component is activated. When the public opinion overview component is activated, the topic sentiment analysis component, the overall trend analysis component, and the public opinion suggestion and risk assessment component are activated.

[0016] Component analysis: When a component is enabled, the corresponding analysis algorithm is called to conduct in-depth analysis of the required data;

[0017] Report generation: Integrate the analysis results of all enabled components to generate public opinion reports;

[0018] Dynamic adjustment: Real-time monitoring of public opinion changes, using a dynamic threshold adjustment algorithm to automatically update the thresholds of various key indicators based on historical data and current public opinion trends.

[0019] In one embodiment, the public opinion heat index includes: H = H1 + H2 + H3 + H4, Among them, H is the total volume, H1 is the number of comments, H2 is the number of reposts, H3 is the number of likes, H4 is the number of other interactions, and H 敏 is the volume of sensitive information, i is the number of discussions on sensitive events;

[0020] When the total volume H is greater than or equal to the total volume threshold T1, the public opinion overview component is enabled and the sensitive information volume H is calculated. 敏 , when the volume of sensitive information H 敏 When the volume is greater than the sensitive information threshold T2, analyze the sensitive content in the event or public opinion.

[0021] In one embodiment, the sentiment index includes: E = (E 正 -E 负 )÷E 总 , E1=(E 正 ÷E 总 )×100%, E2=(E 负 ÷E 总 )×100%, where E is the sentiment score, E 正 is the number of positive articles, E 负 is the number of negative articles, E 总 is the total number of analyzed articles, E1 is the proportion of positive public opinion, and E2 is the proportion of negative public opinion;

[0022] When the public opinion overview component is enabled, the topic sentiment analysis component is enabled and the sentiment score E is calculated. When E>0, the positive public opinion ratio E1 is calculated and positive sentiment analysis is enabled; when E<0, the negative public opinion ratio E2 is calculated and negative sentiment analysis is enabled.

[0023] In one embodiment, the public opinion participation index includes: P = P1 + P2 + P3, where P is user participation, P1 is the number of users participating in the comment, P2 is the number of users forwarding, and P3 is the number of users who like the post;

[0024] When the user participation P is greater than or equal to the public opinion participation threshold T3, the channel analysis component is enabled to analyze the number of participating media and the number of hot topics.

[0025] In one of the embodiments, the account propagation indicator includes the number of propagation accounts;

[0026] When the number of communication accounts is greater than or equal to the communication account number threshold T4, the communication account analysis component is enabled, and the communication accounts are sorted according to their voice values, and each communication account is assigned a value to analyze the contribution of different communication channels to public opinion.

[0027] In one embodiment, the hot word index includes: 1,2,3…… =(N 1,2,3…… ÷N 总 )×100%, where F 1,2,3…… are the frequencies of each popular word, N 1,2,3…… are the number of occurrences of each hot word, N 总 is the total number of words;

[0028] When the frequency of one of the hot words is greater than or equal to the keyword frequency threshold T5, the word cloud analysis component is enabled to generate a visual word cloud of the hot words to display the high-frequency words.

[0029] In one embodiment, the dynamic threshold adjustment algorithm is specifically: T′=T+λ×(Y÷Y 均 -1), where T′ is the adjusted threshold, T is the current threshold, λ is the adjustment coefficient, Y is the key indicator value of the corresponding current public opinion data, and Y 均 It is the average value of the historical data of the corresponding key indicator value.

[0030] A system for automatically interpreting and reporting public opinion events of the present invention comprises:

[0031] The collection module is used to regularly obtain relevant data from news websites, social media and specific industries;

[0032] A preprocessing module, connected to the acquisition module, for performing OCR recognition, format conversion and invalid data cleaning on the data acquired by the acquisition module to form valid text data in a unified format;

[0033] A storage module, connected to the preprocessing module, for converting the text data processed by the preprocessing module into a high-dimensional semantic vector and storing the high-dimensional semantic vector;

[0034] The indicator, component and algorithm definition module is used to define multiple key indicators, corresponding multiple components, corresponding multiple initial thresholds, and the algorithm logic for enabling each component;

[0035] A logic activation module, connected to the storage module and the indicators, components and algorithm definition modules respectively, and connected with a plurality of components, for calculating the high-dimensional semantic vector corresponding to a certain public opinion in the storage module according to the key indicators in the indicators, components and algorithm definition modules, and activating the corresponding components according to the algorithm logic of the indicators, components and algorithm definition modules;

[0036] A report generation module, connected to the logic activation module, for integrating the components activated by the logic activation module into an analysis result, generating a visual display and proposing strategic recommendations;

[0037] The dynamic adjustment module is connected to the report generation module and the indicator, component and algorithm definition module respectively, and is used to monitor the analysis results of the report generation module, and calculate new thresholds based on the historical data of the analysis results and the current corresponding thresholds, and feed them back to the indicator, component and algorithm definition module.

[0038] A device for automatically interpreting and reporting public opinion events of the present invention comprises a memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores the system for automatically interpreting and reporting public opinion events, and the processor executes the method for automatically interpreting and reporting public opinion events by executing the instructions of the system.

[0039] A computer-readable storage medium of the present invention stores the above-mentioned system for automatically interpreting and reporting public opinion events, and the system is suitable for executing the above-mentioned method for automatically interpreting and reporting public opinion events.

[0040] Compared with the prior art, the technical solution of the present invention has the following beneficial effects:

[0041] 1. The present invention defines a series of intelligent analysis components to flexibly respond to changes in public opinion. These components include public opinion overview component, topic sentiment analysis component, channel analysis component, communication account analysis component, word cloud analysis component, overall trend analysis component, and public opinion suggestion and risk assessment component. The system uses comprehensive activation logic to dynamically select corresponding components based on key indicators to ensure that report generation can fully cover all aspects of public opinion;

[0042] 2. The present invention also establishes a dynamic feedback mechanism to enhance the adaptive ability of the system. The system can automatically adjust the activation threshold of each component according to the real-time changes in public opinion, thereby improving the ability to respond to emergencies. In this way, the system can not only better adapt to the specific needs of users, but also improve the accuracy and practicality of analysis, ensuring that users' decisions in public opinion management are more efficient. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 It is a flow chart of a method for automatically interpreting public opinion events and generating reports according to the present invention;

[0044] Figure 2 It is a structural schematic diagram of a system for automatically interpreting and reporting public opinion events according to the present invention. DETAILED DESCRIPTION

[0045] The drawings are only used for illustrative purposes and should not be construed as limitations on this patent. In order to better illustrate this embodiment, some parts of the drawings may be omitted, enlarged, or reduced, and do not represent the size of the actual product. For those skilled in the art, it is understandable that some well-known structures and their descriptions in the drawings may be omitted.

[0046] In the description of the present invention, it should be noted that, unless otherwise clearly specified and limited, the terms "installation" and "connection" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be directly connected or indirectly connected through an intermediate medium. It can be said that the internal connection of two components is connected. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances. The technical solution of the present invention is further explained below in conjunction with the accompanying drawings and embodiments.

[0047] like Figure 1 As shown, a method for automatically interpreting public opinion events and generating reports of the present invention is applied to an artificial intelligence system, and the method includes:

[0048] Data collection: real-time collection of data from news websites, social media and specific industries;

[0049] Data preprocessing: Convert data in different formats into a standard format, remove invalid, duplicate and noisy data, and extract text data from images;

[0050] Data storage: load the pre-trained model, convert text data into high-dimensional semantic vectors and store them;

[0051] Definition of key indicators, including public opinion heat index, sentiment index, public opinion participation index, account communication index and hot word index;

[0052] Component definitions include the public opinion overview component, topic sentiment analysis component, channel analysis component, communication account analysis component, and word cloud analysis component, which correspond to each key indicator. They also include the overall trend analysis component and the public opinion suggestion and risk assessment component.

[0053] Component activation logic: Set initial thresholds for each key indicator except the sentiment indicator. When the corresponding key indicator exceeds the threshold after calculation, the corresponding component is activated. When the public opinion overview component is activated, the topic sentiment analysis component, the overall trend analysis component, and the public opinion suggestion and risk assessment component are activated.

[0054] Component analysis: When a component is enabled, the corresponding analysis algorithm is called to conduct in-depth analysis of the required data;

[0055] Report generation: Integrate the analysis results of all enabled components to generate public opinion reports;

[0056] Dynamic adjustment: Real-time monitoring of public opinion changes, using a dynamic threshold adjustment algorithm to automatically update the thresholds of various key indicators based on historical data and current public opinion trends.

[0057] The method of the present invention defines a series of intelligent analysis components to flexibly respond to changes in public opinion. These components include a public opinion overview component, a topic sentiment analysis component, a channel analysis component, a communication account analysis component, a word cloud analysis component, an overall trend analysis component, and a public opinion suggestion and risk assessment component, and some components correspond one-to-one with key indicators such as public opinion heat index, sentiment index, public opinion participation index, account communication index, and hot word index, and realize their activation logic by presetting the initial threshold. At the same time, the present invention also establishes a dynamic feedback mechanism, which changes the above thresholds by combining historical data and current public opinion trends to continuously enhance the adaptive ability of the system. In this way, the system can not only better adapt to the specific needs of users, but also improve the accuracy and practicality of the analysis, ensuring that users' decisions in public opinion management are more efficient.

[0058] First, the method of the present invention uses an artificial intelligence system to obtain data from news websites, social media and specific industries, obtain data on relevant news reports and comments, collect user discussions, including text, pictures and videos, and pay attention to relevant data sources of specific industries such as finance and technology. For news websites, it mainly uses the Scrapy application framework to build a web crawler, and uses Cron jobs to run crawlers regularly, regularly crawling the latest public opinion-related reports and comments in target websites such as People's Daily Online and Xinhuanet, and storing the crawled data in the El ast icSearch database, and the structure includes title, link, release date, author and content. For social media, API interfaces are generally used, Twitter API and Weibo API are used to collect user comments, forwarding and likes related to specific topics, and keyword lists are set at the same time to monitor discussions on related topics. These data are also stored in the El ast icSearch database, and the fields include user ID, content, release time, comment type, etc. For specific industry data, relevant data in the fields of finance, technology, etc. can also be integrated through API, and the update frequency is to extract the latest market data from API every hour and parse it into the above-mentioned structured data format.

[0059] Then, data preprocessing is performed, and the Tesseract OCR tool is integrated to regularly process pictures and videos contained in social media, extract text information, and store the extracted text in the database to supplement the data set and ensure the comprehensiveness of the analysis. The Pandas library is used in the database to write the ETL process, uniformly convert the crawled data and the data returned by the API into the DataFrame format, write Python scripts, delete invalid and duplicate data, such as advertisements, spam comments, etc., remove texts with a length of less than 5 characters, and use regular expressions to remove HTML tags and special characters at the same time to ensure the neatness of the text data. The processed data is then stored, and the pre-trained BERT model is loaded using the Transformers library of Hugging Face to convert the text data into high-dimensional vectors, so that vectors can be generated in batches before analysis and stored in the Redis cache to increase the speed of subsequent analysis.

[0060] Secondly, define key indicators, components, initial thresholds, and enabling logic. When a component is enabled, the corresponding analysis algorithm can be called to perform component analysis. Specifically:

[0061] Public opinion heat index includes: H = H1 + H2 + H3 + H4, Among them, H is the total volume, H1 is the number of comments, H2 is the number of reposts, H3 is the number of likes, H4 is the number of other interactions, and H 敏is the volume of sensitive information, i is the number of discussions on sensitive events; when the total volume H is greater than or equal to the total volume threshold T1, for example, the initial threshold of T1 is 1000, the public opinion overview component is enabled and the volume of sensitive information H is calculated. 敏 , when the volume of sensitive information H 敏 When it is greater than the sensitive information volume threshold T2, for example, the initial threshold of T2 is 100, the sensitive content in the event or public opinion is analyzed. The public opinion overview component is the first major activation component, that is, a public opinion needs to reach a certain volume before the component should be activated. It mainly provides the total volume, that is, the number of all comments, reposts, likes and other special interactions such as clicks, passers-by, etc., to analyze the overall situation of public opinion, and includes the following positive and negative public opinion analysis when it is activated.

[0062] Emotional indicators include: E = (E 正 -E 负 )÷E 总 , E1=(E 正 ÷E 总 )×100%, E2=(E 负 ÷E 总 )×100%, where E is the sentiment score, E 正 is the number of positive articles, E 负 is the number of negative articles, E 总 is the total number of analyzed articles, E1 is the proportion of positive public opinion, and E2 is the proportion of negative public opinion; when the public opinion overview component is enabled, the topic sentiment analysis component is enabled and the sentiment score E is calculated. When E>0, the positive public opinion proportion E1 is calculated and positive sentiment analysis is enabled; when E<0, the negative public opinion proportion E2 is calculated and negative sentiment analysis is enabled. The topic sentiment analysis component mainly uses the pre-trained sentiment analysis model to output the sentiment score E, and classifies public opinion and related articles into positive, negative and neutral, and analyzes articles with positive or negative public opinion according to the situation to explore potential hot spots.

[0063] The public opinion participation index includes: P = P1 + P2 + P3, where P is user participation, P1 is the number of users participating in the comment, P2 is the number of users forwarding, and P3 is the number of users who like. When the user participation P is greater than or equal to the public opinion participation threshold T3, for example, the initial threshold of T3 is 500, the channel analysis component is enabled to analyze the number of participating media and the number of hot topics. The difference between the public opinion participation index and the public opinion heat index is that the comments, forwarding and likes in the public opinion heat index are calculated by the number of times, while the public opinion participation index is calculated by the number of users, which can avoid maliciously brushing the number of times by a single user, and the activation of the channel analysis component can be combined with the above-mentioned public opinion overview component to analyze the degree of user participation in various channels such as news and social media to analyze their contribution to the public opinion.

[0064] Account communication indicators include the number of communication accounts; when the number of communication accounts is greater than or equal to the communication account number threshold T4, for example, the initial threshold of T4 is 50, the communication account analysis component is enabled, and the communication accounts are sorted according to their voice values, and each communication account is assigned a value to analyze the contribution of different communication channels to public opinion and identify the communication accounts with the greatest influence.

[0065] Hot word indicators include: F 1,2,3…… =(N 1,2,3…… ÷N 总 )×100%, where F 1,2,3…… are the frequencies of each popular word, N 1,2,3…… are the number of occurrences of each hot word, N 总 is the total number of words; when the frequency of one of the hot words is greater than or equal to the keyword frequency threshold T5, for example, the initial threshold of T5 is 10%, the word cloud analysis component is enabled to generate a visual word cloud of the hot words to display high-frequency words. When there are a large number of articles, the TF-IDF algorithm can also be used for estimation, that is, assuming that there is a document with a total of 100 words and a certain word appears 3 times, then the word frequency of the word in the document is 3%. If the word appears in 1,000 documents and the total number of documents is 10,000,000, then its inverse file frequency is lg(10,000,000÷1,000)=4, and the final TF-IDF score is 3%×4=12%.

[0066] In addition, the overall trend analysis component and the public opinion suggestion and risk assessment component are summary components. When the main public opinion overview component is enabled, these two components will comprehensively form corresponding reports based on the activation status and analysis of the above-mentioned other components. For example, the overall trend analysis component displays the time change trend of public opinion and analyzes peaks and troughs. The public opinion suggestion and risk assessment component provides strategic recommendations based on comprehensive analysis of indicators.

[0067] Finally, all components generate reports. Depending on the nature of each component, they generally need to call database aggregation queries to automatically summarize the total volume and sentiment analysis results, and generate time series graphs to show the trend of public opinion changes. Pie charts can be used to show the proportion of public opinion contributions from each channel, etc. The analysis results of all enabled components are integrated to generate structured reports. Matplotlib is used to generate charts, combined with Seaborn to beautify the visual effects, and based on the analysis results, strategic recommendations and risk assessments for specific issues are generated.

[0068] In addition, the system will monitor in real time, that is, use WebSocket technology to achieve real-time data updates, and dynamically adjust the thresholds of key indicators. The dynamic threshold adjustment algorithm is as follows: T′=T+λ×(Y÷Y 均-1), where T′ is the adjusted threshold, T is the current threshold, λ is the adjustment coefficient, for example, 0.5 is selected, which is suitable for scenarios with medium sensitivity requirements and can balance the dependence on historical data and the response to current data changes. Y is the key indicator value of the corresponding current public opinion data, that is, the corresponding current sentiment score, public opinion heat and other indicator values. Y 均 The average value of the historical data of the corresponding key indicator value is taken, and the sentiment score, public opinion heat and other indicator values ​​of the past week are taken. In this way, each time the public opinion report is obtained, the thresholds are constantly changed and substituted from the initial thresholds, which is equivalent to using historical data to train and learn to adjust the thresholds, continuously optimize the report content and presentation method, and improve the clarity of information transmission.

[0069] In summary, multiple intelligent components of the present invention can be dynamically enabled, and the artificial intelligence system will intelligently select and enable relevant components based on real-time data analysis to ensure the comprehensiveness and pertinence of the analysis results. It also has diversified analysis, and different components support in-depth analysis of various public opinion indicators, providing more detailed insights, such as sentiment scores, public opinion heat, and communication account analysis. The present invention also combines the dynamic adjustment of thresholds, through a dynamic threshold adjustment algorithm, to monitor changes in public opinion in real time, timely update the thresholds of each indicator, and improve the response speed to emergencies.

[0070] like Figure 2 As shown, a system for automatically interpreting public opinion events and generating reports of the present invention includes:

[0071] The collection module is used to regularly obtain relevant data from news websites, social media and specific industries;

[0072] The preprocessing module is connected to the acquisition module and is used to perform OCR recognition, format conversion and invalid data cleaning on the data acquired by the acquisition module to form valid text data in a unified format;

[0073] A storage module, connected to the preprocessing module, for converting the text data processed by the preprocessing module into a high-dimensional semantic vector and storing it;

[0074] The indicator, component and algorithm definition module is used to define multiple key indicators, corresponding multiple components, corresponding multiple initial thresholds, and the algorithm logic for enabling each component;

[0075] The logic activation module is connected to the storage module and the indicator, component and algorithm definition module respectively, and is connected to multiple components, and is used to calculate the high-dimensional semantic vector corresponding to a public opinion in the storage module according to the key indicators in the indicator, component and algorithm definition module, and activate the corresponding component according to the algorithm logic of the indicator, component and algorithm definition module;

[0076] The report generation module is connected to the logic enablement module and is used to integrate the analysis results of the components enabled by the logic enablement module, generate visual displays, and make strategic recommendations;

[0077] The dynamic adjustment module is connected to the report generation module and the indicator, component and algorithm definition module respectively. It is used to monitor the analysis results of the report generation module, and calculate new thresholds based on the historical data of the analysis results and the current corresponding thresholds, and feed them back to the indicator, component and algorithm definition modules.

[0078] The system of the present invention mainly applies the above-mentioned method of automatic interpretation and report generation of public opinion events. It uses comprehensive activation logic to dynamically select corresponding components according to key indicators to ensure that report generation can comprehensively cover all aspects of public opinion, and can automatically adjust the activation threshold of each component according to real-time public opinion changes, thereby improving the ability to respond to emergencies. The system completes the above-mentioned method of automatic interpretation and report generation of public opinion events through effective connection and joint action between modules. Its principle is similar to the above-mentioned method and will not be repeated here.

[0079] A device for automatically interpreting and reporting public opinion events of the present invention comprises a memory and a processor, the memory and the processor are connected to each other in communication, the memory stores the above-mentioned system for automatically interpreting and reporting public opinion events, and the processor executes the above-mentioned method for automatically interpreting and reporting public opinion events by executing the instructions of the system. The memory and the processor of the device can be connected via a bus or other means, and the processor can be a central processing unit CPU, a digital signal processor DSP, an application-specific integrated circuit ASIC, a field programmable gate array FPGA or other editable logic device, a discrete gate or transistor logic device, a discrete hardware component, or a combination of the above-mentioned chips. The memory, as a non-transient computer-readable storage medium, can be used to store non-transient software programs, non-transient computer executable programs and modules, such as the program instructions / modules corresponding to the above-mentioned pre-training method. The processor executes various functional applications and data processing of the processor by running the non-transient software programs, instructions and modules stored in the memory, that is, the above-mentioned pre-training method is realized. The memory may include a program storage area and a data storage area, wherein the program storage area can store an operating system and an application required by at least one function; the data storage area can store data created by the processor, etc. In addition, the memory may include a high-speed random access memory, and may also include a non-transitory memory, such as at least one disk storage device, a flash memory device, or other non-transitory solid-state storage device. In some embodiments, the memory may optionally include a memory remotely arranged relative to the processor, and these remote memories may be connected to the processor via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0080] A computer-readable storage medium of the present invention stores a system for automatically interpreting and reporting public opinion events, and the system is suitable for executing the method for automatically interpreting and reporting public opinion events. All or part of the processes in the above-mentioned pre-training method can be completed by instructing related hardware through a computer program, and the implemented program can be stored in a computer-readable storage medium, which, when executed, may include the processes of the above-mentioned methods. Among them, the storage medium can be a magnetic disk, an optical disk, a read-only storage memory ROM, a random access memory RAM, a flash memory, a hard disk HDD or a solid-state drive SSD, etc., or it can also be a combination of the above-mentioned types of memory.

[0081] In the figure, the description of the positional relationship is only for illustrative purposes and should not be understood as a limitation on this patent; obviously, the above embodiments of the present invention are only examples for clearly illustrating the present invention, and are not limitations on the implementation methods of the present invention. For ordinary technicians in the relevant field, other different forms of changes or modifications can be made on the basis of the above description. It is not necessary and impossible to list all the implementation methods here. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the scope of protection of the claims of the present invention.

Claims

1. A method for automatically interpreting public opinion events and generating reports, characterized in that: The method is applied to an artificial intelligence system, and the method comprises: Data collection: real-time collection of data from news websites, social media and specific industries; Data preprocessing: Convert data in different formats into a standard format, remove invalid, duplicate and noisy data, and extract text data from images; Data storage: load the pre-trained model, convert text data into high-dimensional semantic vectors and store them; Definition of key indicators, including public opinion heat index, sentiment index, public opinion participation index, account communication index and hot word index; Component definitions include the public opinion overview component, topic sentiment analysis component, channel analysis component, communication account analysis component, and word cloud analysis component, which correspond to each key indicator. They also include the overall trend analysis component and the public opinion suggestion and risk assessment component. Component activation logic: set initial thresholds for each key indicator other than the sentiment indicator. When the corresponding key indicator exceeds the threshold after calculation, the corresponding component is activated. When the public opinion overview component is activated, the topic sentiment analysis component, the overall trend analysis component, and the public opinion suggestion and risk assessment component are activated. Component analysis: When a component is enabled, the corresponding analysis algorithm is called to conduct in-depth analysis of the required data; Report generation: Integrate the analysis results of all enabled components to generate public opinion reports; Dynamic adjustment: Real-time monitoring of public opinion changes, using a dynamic threshold adjustment algorithm to automatically update the thresholds of various key indicators based on historical data and current public opinion trends.

2. According to the method of automatic interpretation and report generation of public opinion events as described in claim 1, it is characterized in that: The public opinion heat index includes: H = H1 + H2 + H3 + H4, Among them, H is the total volume, H1 is the number of comments, H2 is the number of reposts, H3 is the number of likes, H4 is the number of other interactions, and H 敏 is the volume of sensitive information, i is the number of discussions on sensitive events; When the total volume H is greater than or equal to the total volume threshold T1, the public opinion overview component is enabled and the sensitive information volume H is calculated. 敏 , when the volume of sensitive information H 敏 When the volume is greater than the sensitive information threshold T2, analyze the sensitive content in the event or public opinion.

3. According to the method of automatic interpretation and report generation of public opinion events as described in claim 2, it is characterized in that: The emotional index includes: E = (E 正 -E 负 )÷E 总 , E1=(E 正 ÷E 总 )×100%, E2=(E 负 ÷E 总 )×100%, where E is the sentiment score, E 正 is the number of positive articles, E 负 is the number of negative articles, E 总 is the total number of analyzed articles, E1 is the proportion of positive public opinion, and E2 is the proportion of negative public opinion; When the public opinion overview component is enabled, the topic sentiment analysis component is enabled and the sentiment score E is calculated. When E>0, the positive public opinion ratio E1 is calculated and positive sentiment analysis is enabled; when E<0, the negative public opinion ratio E2 is calculated and negative sentiment analysis is enabled.

4. According to the method of automatic interpretation and report generation of public opinion events as described in claim 1, it is characterized in that: The public opinion participation index includes: P = P1 + P2 + P3, where P is the user participation, P1 is the number of users participating in the comment, P2 is the number of users forwarding, and P3 is the number of users who like the post; When the user participation P is greater than or equal to the public opinion participation threshold T3, the channel analysis component is enabled to analyze the number of participating media and the number of hot topics.

5. According to the method of automatic interpretation and report generation of public opinion events as described in claim 1, it is characterized in that: The account propagation indicator includes the number of propagation accounts; When the number of communication accounts is greater than or equal to the communication account number threshold T4, the communication account analysis component is enabled, and the communication accounts are sorted according to their voice values, and each communication account is assigned a value to analyze the contribution of different communication channels to public opinion.

6. According to the method of automatic interpretation and report generation of public opinion events as described in claim 1, it is characterized in that: The hot word indicators include: 1,2,3…… =(N 1,2,3…… ÷N 总 )×100%, where F 1,2,3…… are the frequencies of each popular word, N 1,2,3…… are the number of occurrences of each hot word, N 总 is the total number of words; When the frequency of one of the hot words is greater than or equal to the keyword frequency threshold T5, the word cloud analysis component is enabled to generate a visual word cloud of the hot words to display the high-frequency words.

7. According to claim 1, a method for automatically interpreting and reporting public opinion events is characterized in that: The dynamic threshold adjustment algorithm is specifically: T′=T+λ×(Y÷Y 均 -1), where T′ is the adjusted threshold, T is the current threshold, λ is the adjustment coefficient, Y is the key indicator value of the corresponding current public opinion data, and Y 均 It is the average value of the historical data of the corresponding key indicator value.

8. A system for automatically interpreting public opinion events and generating reports, characterized in that: include: The collection module is used to regularly obtain relevant data from news websites, social media and specific industries; A preprocessing module, connected to the acquisition module, for performing OCR recognition, format conversion and invalid data cleaning on the data acquired by the acquisition module to form valid text data in a unified format; A storage module, connected to the preprocessing module, for converting the text data processed by the preprocessing module into a high-dimensional semantic vector and storing the high-dimensional semantic vector; The indicator, component and algorithm definition module is used to define multiple key indicators, corresponding multiple components, corresponding multiple initial thresholds, and the algorithm logic for enabling each component; A logic activation module, connected to the storage module and the indicators, components and algorithm definition modules respectively, and connected with a plurality of components, for calculating the high-dimensional semantic vector corresponding to a certain public opinion in the storage module according to the key indicators in the indicators, components and algorithm definition modules, and activating the corresponding components according to the algorithm logic of the indicators, components and algorithm definition modules; A report generation module, connected to the logic activation module, for integrating the components activated by the logic activation module into an analysis result, generating a visual display and proposing strategic recommendations; The dynamic adjustment module is connected to the report generation module and the indicator, component and algorithm definition module respectively, and is used to monitor the analysis results of the report generation module, and calculate new thresholds based on the historical data of the analysis results and the current corresponding thresholds, and feed them back to the indicator, component and algorithm definition module.

9. A device for automatically interpreting public opinion events and generating reports, characterized in that: It includes a memory and a processor, the memory and the processor are communicatively connected to each other, the memory stores a system for automatically interpreting and reporting public opinion events as described in claim 8, and the processor executes the method for automatically interpreting and reporting public opinion events as described in any one of claims 1-7 by executing instructions of the system.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a system for automatically interpreting and reporting public opinion events as described in claim 8, and the system is suitable for executing a method for automatically interpreting and reporting public opinion events as described in any one of claims 1-7.

Images