Public opinion monitoring and risk early warning system and method based on investor interactive questions and answers

Abstract

The invention provides a public opinion monitoring and risk early warning system and method based on investor interactive questions and answers, and the system comprises a collection and preprocessing module which is used for collecting and preprocessing historical investor and listed company interactive questions and answers data, real-time investor and listed company interactive questions and answers data and a third-party data source; the data processing module is used for performing data processing on the real-time interactive question and answer data of the investors and the listed companies; and the public opinion wind score calculation module is used for calculating a public opinion risk score through a public opinion risk score calculation model and sending an early warning notification to the user according to the public opinion risk score. According to the method, the whole process of collection, analysis, early warning and display is automated, the public opinion analysis time is shortened from the hour level to the minute level or even the second level, and manual intervention is reduced.

Description

Technical Field

[0001] This application relates to the technical field of public opinion monitoring, and specifically to a public opinion monitoring and risk early warning system and method based on investor interactive questions and answers. Background Art

[0002] Currently, there are many mature technologies in the fields of sentiment analysis, real-time monitoring, and risk warning. These technologies have been widely applied in areas such as stock markets, cryptocurrency trading, and investment decision support. However, existing technologies primarily focus on public social media or news, rather than specifically targeting the specific scenario of "interactive Q&A between investors and listed companies." Furthermore, existing technologies tend to focus more on macro news or announcements, rather than directly analyzing real-time public opinion during investor Q&A sessions.

[0003] However, the calculation of public opinion monitoring risk scores in the existing technology does not take into account the interaction effects between emotional risk indicators, avoidance indicators and individual indicators, and the calculation of public opinion monitoring risk scores is inaccurate and incomplete. Summary of the Invention

[0004] In order to help solve the above technical problems, this application provides a public opinion monitoring and risk warning system and method based on investor interactive questions and answers.

[0005] First, this application provides a method for public opinion monitoring and risk warning based on investor interactive questions and answers, which adopts the following technical solutions:

[0006] A public opinion monitoring and risk early warning system based on investor interactive question-and-answer sessions, wherein the system comprises:

[0007] The collection and preprocessing module is used to collect and preprocess historical investor and listed company interactive question and answer data, real-time investor and listed company interactive question and answer data, and third-party data sources;

[0008] Data processing module, including:

[0009] A first data processing unit is used to segment the interactive question-and-answer data between real-time investors and listed companies using a basic word segmentation tool, extract sensitive words using a pre-trained sensitive word extraction model and algorithm, and calculate sensitive content indicators;

[0010] a second data processing unit, configured to perform sentiment analysis on the interactive question-and-answer data between real-time investors and listed companies using a preset text classification model to obtain sentiment data, calculate a sentiment index based on the sentiment data, and calculate an avoidance index for the interactive question-and-answer data between real-time investors and listed companies based on a predefined avoidance expression;

[0011] a third data processing unit, configured to classify the interactive question-and-answer data between real-time investors and listed companies using a preset classifier, and identify abnormal question patterns in the interactive question-and-answer data between real-time investors and listed companies using a preset recognition algorithm, wherein the abnormal question patterns include batch repeated questions and concentrated negative questions;

[0012] The public opinion risk score calculation module is used to calculate the public opinion risk score based on the sensitive content index, sentiment index, avoidance index, and response speed index through the public opinion risk score calculation model, and send early warning notifications to users based on the public opinion risk score.

[0013] Preferably, the public opinion risk score calculation module calculates the public opinion risk score in the following manner:

[0014] Public opinion risk score = wS + wC + wE + wI + w5R + w6 (S × C) + w7 (E × I), where R represents the response speed index, S represents the sentiment risk index, I represents the investor satisfaction index, E represents the avoidance index, and C represents the sensitive content index. S × C represents the interaction effect between the sentiment risk index and the sensitive content index, E × I represents the interaction effect between the avoidance index and the investor satisfaction index, and w, w, w, w, w5, w6, and w7 represent the corresponding weights.

[0015] Preferably, in the first data processing unit, the sensitive content indicator is the proportion of sensitive words in the real-time interactive question and answer data between investors and listed companies.

[0016] Preferably, the system further includes a report module, the sentiment data includes a sentiment score and a satisfaction score, the sentiment index includes a sentiment risk index and an investor satisfaction index, and the second data processing unit calculates the sentiment risk index in the following manner:

[0017] S = 1-((emotion score + 1) / 2), where S represents the emotional risk index;

[0018] The investor satisfaction index is calculated as follows:

[0019] I = 1-((satisfaction score + 1) / 2), where I represents the investor satisfaction index.

[0020] In a second aspect, the present application provides a method for a public opinion monitoring and risk warning system based on investor interactive question-and-answer as described in any one of the first aspects above, which adopts the following technical solutions:

[0021] A method for public opinion monitoring and risk early warning based on investor interactive question and answer, wherein the method comprises:

[0022] Step S1: collecting and preprocessing historical investor and listed company interactive question and answer data, real-time investor and listed company interactive question and answer data, and third-party data sources through the collection and preprocessing module;

[0023] Step S2: Processing the real-time investor and listed company interactive question and answer data through the data processing module, including:

[0024] Step S21: The first data processing unit performs word segmentation on the interactive question and answer data of real-time investors and listed companies based on a basic word segmentation tool, extracts sensitive words using a pre-trained sensitive word extraction model and sensitive word extraction algorithm, and calculates sensitive content indicators;

[0025] Step S22: The second data processing unit performs sentiment analysis on the interactive question-and-answer data between real-time investors and listed companies based on a preset text classification model to obtain sentiment data, calculates a sentiment index based on the sentiment data, and calculates an avoidance index for the interactive question-and-answer data between real-time investors and listed companies based on a predefined avoidance expression;

[0026] Step S23: Classifying the interactive question-and-answer data between real-time investors and listed companies based on a preset classifier by the third data processing unit, and identifying abnormal question patterns in the interactive question-and-answer data between real-time investors and listed companies using a preset recognition algorithm, wherein the abnormal question patterns include batch repeated questions and concentrated negative questions;

[0027] Step S3: The public opinion risk score calculation module calculates the public opinion risk score based on the public opinion risk score calculation model according to the sensitive content index, sentiment index, avoidance index, and response speed index in step S2, and sends an early warning notification to the user based on the public opinion risk score.

[0028] Preferably, the step S1 includes:

[0029] Step S11: Import historical investor and listed company interactive question and answer data, and train and establish an initial public opinion monitoring model through the historical investor and listed company interactive question and answer data;

[0030] Step S12: Introduce a message queue and establish an event-driven data collection service through the message queue. When investors and listed companies interact with each other for questions and answers, the analysis process is immediately triggered;

[0031] Step S13: Access the company's internal third-party data source and associate the third-party data source with the interactive question and answer data of investors and listed companies and the third-party data.

[0032] Preferably, the response speed index is calculated as follows:

[0033] R = min(1, actual response time / standard response time), where R represents the response speed indicator.

[0034] Preferably, step S3 includes:

[0035] Step S31: Calculate the public opinion risk score based on the sensitive content index, sentiment data, avoidance index, and response speed index:

[0036] Step S32: Determine the risk warning level based on the public opinion risk score;

[0037] Step S33: Send an early warning notification to the designated user according to the risk warning level.

[0038] Preferably, step S3 includes: setting a number of risk thresholds, and determining the risk warning level according to the numerical relationship between the public opinion risk score and the risk threshold.

[0039] Preferably, the method further includes step S4, which is executed by the reporting module, comprising:

[0040] S41: Set up a real-time risk monitoring report to display the public opinion risk score and risk warning level of listed companies;

[0041] S42: Set up multi-dimensional data analysis reports to display changes in sentiment trends, including the distribution ratios of positive, negative, and neutral sentiments, and present hot issues and sensitive words that investors are concerned about;

[0042] S43: Set up an abnormal question detection report to highlight abnormally negative or sensitive questions and answers, automatically identify and mark answers containing sensitive words or highly evasive questions, and display different types of abnormal questions in a classified manner.

[0043] To sum up, compared with traditional solutions, this application has better performance, supports high concurrency and low latency, is more powerful in functions such as multi-dimensional analysis and risk prediction, and is suitable for the financial field.

[0044] The advantages of this application include but are not limited to the following aspects:

[0045] 1. Improve the efficiency of public opinion monitoring and data processing

[0046] Compared with traditional manual monitoring, the entire process of collection, analysis, early warning and display is automated, shortening the public opinion analysis time from hours to minutes or even seconds, reducing manual intervention. At the same time, preset models and algorithms are used to process public opinion data, improving data processing efficiency.

[0047] 2. Enhance risk prevention and control capabilities

[0048] Through multi-dimensional anomaly detection and avoidance index calculation, it provides more sensitive risk perception, identifies risk signals such as evasive answers and concentrated negative questions in advance, and reduces the probability of public opinion crises.

[0049] 3. Support accurate decision-making

[0050] Operations staff can quickly develop response strategies based on real-time data and trend analysis, such as adjusting audit strategies or focusing on specific companies.

[0051] 4. Adaptability to financial scenarios

[0052] The system uses an exclusive interactive Q&A data source and integrates it with third-party data. It performs well in public opinion monitoring in the financial field, and can correlate stock market fluctuations with Q&A content to reveal potential market impacts. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 This is a flow chart of an embodiment of a method for public opinion monitoring and risk warning based on investor interactive question and answer in this application. DETAILED DESCRIPTION

[0054] The present invention will be further described below with reference to the accompanying drawings. The structure and principle of the present invention will be very clear to those skilled in the art. It should be understood that the specific embodiments described herein are only intended to explain the present invention and are not intended to limit the present invention.

[0055] This application proposes a public opinion monitoring and risk warning system based on investor interactive questions and answers, which includes an acquisition and preprocessing module, a data processing module, a public opinion risk score calculation module, and a reporting module. The acquisition and preprocessing module is used to acquire and preprocess historical investor and listed company interactive question and answer data, real-time investor and listed company interactive question and answer data, and third-party data sources; the data processing module includes a first data processing unit, a second data processing unit, and a third data processing unit. The first data processing unit is used to segment the interactive question and answer data of real-time investors and listed companies using a basic word segmentation tool, extract sensitive words using a pre-trained sensitive word extraction model and a sensitive word extraction algorithm, and calculate sensitive content indicators; the second data processing unit is used to perform sentiment analysis on the interactive question and answer data of real-time investors and listed companies using a preset text classification model, obtain sentiment data, and calculate sentiment indicators based on the sentiment data. According to the predefined avoidance expression, the avoidance index of the interactive question and answer data of real-time investors and listed companies is calculated. The third data processing unit is used to classify the interactive question and answer data between real-time investors and listed companies through a preset classifier, and to identify abnormal questioning patterns in the interactive question and answer data between real-time investors and listed companies through a preset recognition algorithm. The abnormal questioning patterns include batch repeated questions and concentrated negative questions.

[0056] The public opinion risk score calculation module is used to calculate the public opinion risk score based on the sensitive content index, sentiment index, avoidance index, and response speed index through the public opinion risk score calculation model, and send early warning notifications to users based on the public opinion risk score.

[0057] By comprehensively collecting all relevant data, utilizing various data processing units to conduct multi-dimensional analysis of real-time interactive Q&A data, extracting key characteristic indicators, and then comprehensively assessing public opinion risks through a public opinion risk scoring model, timely warnings are issued, and this enables monitoring and risk warning of public opinion surrounding investor Q&A interactions. By integrating multiple data sources, we can more comprehensively understand the interactive Q&A sessions between investors and listed companies and promptly identify potential public opinion risks.

[0058] The public opinion risk score calculation module calculates the public opinion risk score in the following way:

[0059] Public opinion risk score = wS + wC + wE + wI + w5R + w6 (S × C) + w7 (E × I), where R represents the response speed index, S represents the sentiment risk index, I represents the investor satisfaction index, E represents the avoidance index, and C represents the sensitive content index. S × C represents the interaction effect between the sentiment risk index and the sensitive content index, E × I represents the interaction effect between the avoidance index and the investor satisfaction index, and w, w, w, w, w5, w6, and w7 represent the corresponding weights.

[0060] This formula comprehensively considers multiple factors influencing public opinion risk, including not only the independent impact of each indicator (S, C, E, I, R) but also the interactive effects between the sentiment risk indicator and the sensitive content indicator, and the avoidance indicator and the investor satisfaction indicator (S×C, E×I). By assigning different weights (w to w7), the different importance of each factor to public opinion risk is reflected. This allows for a more comprehensive and accurate assessment of public opinion risk, considering not only the impact of individual indicators but also the interactions between them, making the assessment more realistic.

[0061] In the first data processing unit, the sensitive content indicator is the proportion of sensitive words in the real-time interactive question-and-answer data between investors and listed companies. Sentiment data includes sentiment scores and satisfaction scores, and sentiment indicators include sentiment risk indicators and investor satisfaction indicators. The second data processing unit calculates the sentiment risk indicator through the following method:

[0062] S = 1-((emotion score + 1) / 2), where S represents the emotional risk index;

[0063] The investor satisfaction index is calculated as follows:

[0064] I = 1-((satisfaction score + 1) / 2), where I represents the investor satisfaction index.

[0065] By counting the number of sensitive words in real-time interactive Q&A data and calculating their proportion of the total number of words, we measure the proportion of sensitive content in the data. This serves as a sensitive content indicator, reflecting the degree of sensitive information in the data. Quantifying sensitive content as a proportion facilitates integrated calculation and analysis with other indicators, improving the scientific nature and objectivity of public opinion risk assessments.

[0066] Next, the method section of this application is introduced.

[0067] Figure 1 This is a flow chart of an embodiment of a method for public opinion monitoring and risk early warning based on investor interactive question and answer in this application, which includes:

[0068] Step S1: Collect and pre-process historical investor and listed company interactive question and answer data, real-time investor and listed company interactive question and answer data, and third-party data sources, specifically including:

[0069] Step S11: Import historical interactive question and answer data between investors and listed companies, and establish an initial public opinion monitoring model through training with the historical interactive question and answer data between investors and listed companies, which is used for initial review of the interactive question and answer data between investors and listed companies.

[0070] Step S12: Introduce a message queue and establish an event-driven data collection service through the message queue. When investors and listed companies interact with each other for questions and answers, the analysis process is immediately triggered.

[0071] Step S13: Access the company's internal third-party data source and associate the third-party data source with the interactive question and answer data of investors and listed companies and the third-party data.

[0072] Step S1 primarily involves iterative judgment, with the introduction of third-party data serving as a reference. Public opinion monitoring is primarily used by internal auditors. In practical interactive communication platforms, both investor questions and listed companies' responses require manual backend review before they can be posted online. This often leads to complaints, reports, and malicious inquiries that adversely impact backend review efforts. To proactively address these issues, facilitate backend review efforts, and mitigate public opinion risks, a public opinion monitoring model has been developed.

[0073] Specifically, batch import CSV file data of historical questions and answers to establish an initial model and baseline for public opinion monitoring; introduce message queues such as MQ to establish an event-driven data collection service to immediately trigger the analysis process when investors ask questions and listed companies answer; access the company's internal third-party data sources, such as listed companies' stock transaction summary data, index market data, etc., associate the question and answer data with these third-party data to improve the analysis perspective.

[0074] The initial model is trained through historical data to provide a basis for subsequent public opinion monitoring; real-time data collection is achieved through message queues, and the analysis process is immediately triggered when interactive questions and answers occur to ensure the timeliness of the data; access to third-party data sources and correlation of data enriches the data dimension and improves the comprehensiveness of public opinion monitoring.

[0075] The subsequent step, S2, involves incremental changes. The system's accumulated historical Q&A data from recent years, combined with the third-party data introduced in step S1, serves as the initial training for the model. Step S2 involves the newly generated Q&A data from daily review processes. This data, as a new data stream, also needs to be reflected in the model. Overall, this can be understood as a continuous iterative process.

[0076] Step S2: Analyze the real-time interactive Q&A data between investors and listed companies, including:

[0077] Step S21: Segment the interactive question and answer data between real-time investors and listed companies using a basic word segmentation tool (jieba word segmentation tool), extract sensitive words using a pre-trained sensitive word extraction model (DistilBERT model) and a sensitive word extraction algorithm (TF-IDF algorithm), and calculate the sensitive content index, which is the proportion of sensitive words in the interactive question and answer data between real-time investors and listed companies.

[0078] Specifically, the sensitive word detection module of the proprietary system is used to filter out illegal content based on dictionary matching and context analysis, and the investor and the company being questioned are marked; for keyword extraction, the text is first segmented using basic word segmentation tools to extract a list of keywords related to finance, market, and market conditions. At the same time, the question-answer corresponding data is obtained through the background database to analyze the semantic consistency between the question and answer and the satisfaction of the answer.

[0079] Step S22: sentiment analysis is performed on the interactive question and answer data between real-time investors and listed companies through a preset text classification model (sentiment-analysis model) under the Transformers framework to obtain sentiment data, and sentiment indicators are calculated based on the sentiment data. According to the predefined avoidance expression, the avoidance index of the interactive question and answer data between real-time investors and listed companies is calculated. In the embodiment of the present application, the avoidance expression can be "please pay attention", "not convenient to disclose at the moment", "the specific situation needs further observation", "no specific information", "will be disclosed in the future", etc., and the avoidance index can be the frequency of the avoidance expression, etc. Sentiment data includes sentiment scores and satisfaction scores, and sentiment indicators include sentiment risk indicators and investor satisfaction indicators. The sentiment risk index is calculated in the following way:

[0080] S = 1-((emotion score + 1) / 2), where S represents the emotional risk index, ranging from [0, 1]. Higher values indicate stronger negative emotions.

[0081] The investor satisfaction index is calculated as follows:

[0082] I = 1-((satisfaction score + 1) / 2), where I represents the investor satisfaction index.

[0083] Specifically, sentiment analysis is conducted on questions and answers based on a predefined list of sensitive words, evasive expression patterns, and basic indicators; the evasiveness index of the answer content is calculated. The larger the index data, the more evasive it is, and potential problems that may be implied by vague expressions, evasive language, etc. in the company's answers are identified.

[0084] Step S23: Classify the real-time interactive Q&A data between investors and listed companies using a preset classifier (BERT classifier). Use a preset recognition algorithm (isolation forest algorithm) to identify abnormal question patterns in this data. These abnormal question patterns include batches of repetitive questions and concentrated negative questions. Step S23 is a preprocessing step preceding Step S3.

[0085] Step S3: Calculate the public opinion risk score based on the sensitive content index, sentiment index, avoidance index, and response speed index in step S2 through the public opinion risk score calculation model, and send an early warning notification to the designated user based on the public opinion risk score.

[0086] Step S3 includes: the public opinion risk score calculation model calculates the public opinion risk score in the following manner:

[0087] Public Opinion Risk Score = wS + wC + wE + wI + w5R + w6 (S × C) + w7 (E × I). R represents the response speed indicator, S represents the sentiment risk indicator, I represents the investor satisfaction indicator (range: [0, 1], with higher values indicating lower satisfaction), E represents the avoidance indicator (range: [0, 1], with higher values indicating greater avoidance), and C represents the sensitive content indicator (range: [0, 1], with higher values indicating more sensitive content). S × C represents the interaction effect between the sentiment risk indicator and the sensitive content indicator, and E × I represents the interaction effect between the avoidance indicator and the investor satisfaction indicator. w, w, w, w, w5, w6, and w7 represent the corresponding weights. The response speed indicator R can be calculated based on the question and answer time when pulling listed company response data from the backend. This response speed indicator is calculated based on some initially defined timeliness indicators.

[0088] The response speed index is calculated as follows:

[0089] R = min(1, actual response time / standard response time), where R represents the response speed indicator and ranges from [0 to 1]. Higher values indicate slower responses. The standard response time is a preset value that reflects how long it typically takes a company to respond to a question. Actual response time is calculated as follows: time it takes the listed company to respond - time it takes the investor to ask the question = time it takes to respond.

[0090] Step S3 includes:

[0091] Step S31: Calculate the public opinion risk score based on the sensitive content index, sentiment data, avoidance index and response speed index.

[0092] Step S32: Determine the risk warning level based on the public opinion risk score. In this embodiment, several risk thresholds are set, and the risk warning level is determined based on the numerical relationship between the public opinion risk score and the risk threshold.

[0093] Step S33: Send an early warning notification to the designated user according to the risk warning level.

[0094] Step S4 includes:

[0095] S41: Set up a real-time risk monitoring report to display the public opinion risk score and risk warning level of listed companies. Specifically, the real-time risk monitoring report can intuitively display the public opinion risk index and warning level of listed companies, quickly identify the risk level through a color coding system (green, yellow, orange, red), track the trend of public opinion changes in real time, and promptly detect abnormal fluctuations.

[0096] S42: Create a multi-dimensional data analysis report to demonstrate sentiment trends, including the distribution ratios of positive, negative, and neutral sentiment, and to present investor-focused hot topics and keywords. Specifically, the report will demonstrate sentiment trends, including the distribution ratios of positive, negative, and neutral sentiment, and highlight investor-focused hot topics and keywords. It will also analyze the quality of responses, including metrics such as avoidance, satisfaction, and response speed.

[0097] S43: Set up an abnormal question detection report to highlight abnormally negative or sensitive Q&A content, automatically identify and mark answers containing sensitive words or highly evasive responses, and categorize and display different types of abnormal questions. Specifically, the abnormal question detection report highlights abnormally negative or sensitive Q&A content, automatically identify and mark answers containing sensitive words or highly evasive responses, and categorize and display different types of abnormal questions, such as negative questions and evasive responses.

[0098] Default weight settings

[0099] w(emotional risk weight) = 0.35

[0100] w (sensitive content weight) = 0.15

[0101] w (avoidance weight) = 0.15

[0102] w (satisfaction weight) = 0.15

[0103] w5 (response speed weight) = 0.10

[0104] w6 (emotion-sensitivity interaction weight) = 0.05

[0105] w7 (avoidance-satisfaction interaction weight) = 0.05

[0106] Risk warning levels are divided into serious, severe, and mild, and the specific settings are as follows:

[0107] "severe": 0.7 - Severe risk threshold. When the calculated risk index reaches or exceeds 0.7, the system triggers a severe (SEVERE) alert. This indicates that the company is facing a highly negative public opinion environment and may require immediate intervention.

[0108] "moderate": 0.5 - Moderate risk threshold. When the risk index is between 0.5 and 0.7, the system triggers a moderate-level (MODERATE) warning. This indicates a clear public opinion issue that requires close attention.

[0109] "mild": 0.3 - Mild risk threshold. When the risk index is between 0.3 and 0.5, the system triggers a mild (MILD) alert. This indicates that there are some public opinion risks that require attention but are not yet serious.

Claims

1. A public opinion monitoring and risk early warning system based on investor interactive question and answer, characterized by: The system comprises: The collection and preprocessing module is used to collect and preprocess real-time investor and listed company interactive question and answer data; the data processing module is used to process the real-time investor and listed company interactive question and answer data, including: A first data processing unit is used to segment the interactive question-and-answer data between real-time investors and listed companies using a basic word segmentation tool, extract sensitive words using a pre-trained sensitive word extraction model and algorithm, and calculate sensitive content indicators; a second data processing unit, configured to perform sentiment analysis on the interactive question-and-answer data between real-time investors and listed companies using a preset text classification model to obtain sentiment data, calculate a sentiment index based on the sentiment data, and calculate an avoidance index for the interactive question-and-answer data between real-time investors and listed companies based on a predefined avoidance expression; The third data processing unit is used to classify the interactive question and answer data between real-time investors and listed companies through a preset classifier, and identify abnormal questioning patterns in the interactive question and answer data between real-time investors and listed companies through a preset recognition algorithm. The abnormal questioning patterns include batch repeated questions and concentrated negative questions; the public opinion risk score calculation module is used to calculate the public opinion risk score based on the sensitive content index, sentiment index, avoidance index, and response speed index through the public opinion risk score calculation model, and send early warning notifications to users based on the public opinion risk score.

2. The public opinion monitoring and risk early warning system based on investor interactive question and answer according to claim 1 is characterized in that: The public opinion risk score calculation module calculates the public opinion risk score in the following way: Public opinion risk score = wS + wC + wE + wI + w5R + w6 (S × C) + w7 (E × I), where R represents the response speed index, S represents the sentiment risk index, I represents the investor satisfaction index, E represents the avoidance index, and C represents the sensitive content index. S × C represents the interaction effect between the sentiment risk index and the sensitive content index, E × I represents the interaction effect between the avoidance index and the investor satisfaction index, and w, w, w, w, w5, w6, and w7 represent the corresponding weights.

3. The public opinion monitoring and risk early warning system based on investor interactive question and answer according to claim 1 is characterized in that: In the first data processing unit, the sensitive content indicator is the proportion of sensitive words in the real-time interactive question and answer data between investors and listed companies.

4. The public opinion monitoring and risk early warning system based on investor interactive question and answer according to claim 1 is characterized in that: The system further includes a report module, the sentiment data includes a sentiment score and a satisfaction score, the sentiment index includes a sentiment risk index and an investor satisfaction index, and the second data processing unit calculates the sentiment risk index by: S = 1-((emotion score + 1) / 2), where S represents the emotional risk index; The investor satisfaction index is calculated as follows: I = 1-((satisfaction score + 1) / 2), where I represents the investor satisfaction index.

5. A method for using the public opinion monitoring and risk early warning system based on investor interactive question and answer as described in any one of claims 1 to 4, characterized in that: The method comprises: Step S1: collecting and preprocessing historical investor and listed company interactive question and answer data, real-time investor and listed company interactive question and answer data, and third-party data sources through the collection and preprocessing module; Step S2: Processing the real-time investor and listed company interactive question and answer data through the data processing module, including: Step S21: The first data processing unit performs word segmentation on the interactive question and answer data of real-time investors and listed companies based on a basic word segmentation tool, extracts sensitive words using a pre-trained sensitive word extraction model and sensitive word extraction algorithm, and calculates sensitive content indicators; Step S22: The second data processing unit performs sentiment analysis on the interactive question-and-answer data between real-time investors and listed companies based on a preset text classification model to obtain sentiment data, calculates a sentiment index based on the sentiment data, and calculates an avoidance index for the interactive question-and-answer data between real-time investors and listed companies based on a predefined avoidance expression; Step S23: Classifying the interactive question-and-answer data between real-time investors and listed companies based on a preset classifier by the third data processing unit, and identifying abnormal question patterns in the interactive question-and-answer data between real-time investors and listed companies using a preset recognition algorithm, wherein the abnormal question patterns include batch repeated questions and concentrated negative questions; Step S3: The public opinion risk score calculation module calculates the public opinion risk score based on the public opinion risk score calculation model according to the sensitive content index, sentiment index, avoidance index, and response speed index in step S2, and sends an early warning notification to the user based on the public opinion risk score.

6. The method according to claim 1, characterized in that The step S1 comprises: Step S11: Import historical investor and listed company interactive question and answer data, and train and establish an initial public opinion monitoring model through the historical investor and listed company interactive question and answer data; Step S12: Introduce a message queue and establish an event-driven data collection service through the message queue. When investors and listed companies interact with each other for questions and answers, the analysis process is immediately triggered; Step S13: Access the company's internal third-party data source and associate the third-party data source with the interactive question and answer data of investors and listed companies and the third-party data.

7. The method according to claim 1, characterized in that The response speed index is calculated as follows: R = min(1, actual response time / standard response time), where R represents the response speed index.

8. The method according to claim 1, characterized in that The step S3 comprises: Step S31: Calculate the public opinion risk score based on the sensitive content index, sentiment data, avoidance index, and response speed index: Step S32: Determine the risk warning level based on the public opinion risk score; Step S33: Send an early warning notification to the designated user according to the risk warning level.

9. The method according to claim 5, characterized in that The step S3 includes: setting a number of risk thresholds, and determining the risk warning level according to the numerical relationship between the public opinion risk score and the risk threshold.

10. The method according to claim 1, characterized in that The method further includes step S4 for executing the report module, comprising: S41: Set up a real-time risk monitoring report to display the public opinion risk score and risk warning level of listed companies; S42: Set up multi-dimensional data analysis reports to display changes in sentiment trends, including the distribution ratios of positive, negative, and neutral sentiments, and present hot issues and sensitive words that investors are concerned about; S43: Set up an abnormal question detection report to highlight abnormally negative or sensitive questions and answers, automatically identify and mark answers containing sensitive words or highly evasive questions, and display different types of abnormal questions in a classified manner.

Images