Internet hot news intelligent acquisition and management system

Abstract

The invention relates to the technical field of information management, in particular to an Internet hot news intelligent acquisition and management system, which comprises a news acquisition module, a news management module and a news management module, the news labeling module is used for primarily processing the collected news information, adding labels and verifying the accuracy of the news information; the influence analysis module is used for obtaining the comment number, the like number, the reply number, the sharing frequency and the reading amount corresponding to the news information and evaluating the influence analysis coefficient of the news information; identifying the development trend of the news information according to the change condition of the news information along with time; the emotion analysis module is used for identifying positive, negative and neutral emotions in the news, setting emotion scores for the identified emotions, and identifying emotion influence coefficients corresponding to the news information according to the obtained emotion scores; and the news processing efficiency and accuracy are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of information management, in particular to an intelligent collection and management system for hot Internet news. Background Art

[0002] With the development of the Internet, a large amount of news is reported in the form of online text. Efficient mining of online news is an important requirement for economic and social development. Predicting the future direction of a news event is an important and difficult problem with great economic and social value. Existing methods usually rely on domain experts to make inferences based on their own experience and predict the development of subsequent events. However, due to the different backgrounds and inconsistent views of each person, the prediction results often vary greatly, and the accuracy cannot be guaranteed.

[0003] For example, Chinese patent publication number CN113159363A discloses a method for predicting event trends based on historical news reports. First, the field of new events to be predicted is determined, and similar events are obtained based on public data sets or data collected from the Internet; within each similar event, news describing the same specific event is clustered into a sub-event, and the subsequent event distribution information of each sub-event is obtained based on the relevance of the event content and the chronological order; the similarity between similar events and the new event to be predicted is calculated to obtain similar events; then, within similar events, the similarity between the current sub-event and the sub-events in the similar events is obtained by calculating the similarity, and the development trend of the current sub-event is predicted based on the aforementioned two similarities and event distribution information.

[0004] However, when describing and summarizing the time corresponding to the news information, the impact and emotion of the news event itself are ignored. It is necessary to adjust the way of handling news according to the emotions and impacts corresponding to the news to improve the management effect of news and improve the accuracy of news recognition. Summary of the invention

[0005] In order to solve the above technical problems, the technical solution adopted by the present invention is: an intelligent collection and management system for hot Internet news, including: a news collection module for collecting news information from different sources.

[0006] The news annotation module is used to perform preliminary processing on the collected news information, add labels and verify the accuracy of the news information.

[0007] The influence analysis module is used to obtain the number of comments, likes, replies, shares, and readings corresponding to the news information, and evaluate the impact analysis coefficient of the news information; based on the changes in the news information over time, identify the development trend of the news information.

[0008] The sentiment analysis module is used to identify positive, negative and neutral emotions in news, set sentiment scores for the identified emotions, and identify the sentiment impact coefficient corresponding to the news information based on the obtained sentiment scores.

[0009] The event prediction module is used to identify the prediction labels of news information, and classify the news information according to the prediction labels to obtain the comprehensive scores of events corresponding to the news information.

[0010] The comprehensive evaluation module is used to integrate the output results of the influence analysis module, sentiment analysis module and event prediction module to give the final evaluation score.

[0011] The process control module is used to obtain the process control parameters corresponding to the news information according to the final evaluation score and adjust the importance and management strategy of the news.

[0012] The beneficial effects of the present invention are: 1. By calculating the impact analysis coefficient of multi-dimensional data including the number of comments, the number of likes, the number of replies, the number of shares, the number of readings, etc., the system can accurately evaluate the influence of news information and its development trend, providing a scientific basis for judging the news value.

[0013] 2. The present invention can intelligently identify positive, negative and neutral emotions in news and calculate the emotional impact coefficient, helping users understand the public's emotional attitude towards news and providing strong support for public opinion monitoring.

[0014] 3. By predicting the event tags of news information and calculating the comprehensive scores of events, the system can predict and comprehensively evaluate the development trends of news events, providing journalists with forward-looking decision-making references.

[0015] Fourth, the present invention can automatically adjust the importance and management strategies of news according to the final evaluation score, such as release priority, monitoring frequency and feedback response speed, thereby realizing the personalization and intelligence of news management. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The present invention is further described below in conjunction with the accompanying drawings and embodiments.

[0017] Figure 1 It is a system framework diagram of an intelligent collection and management system for Internet hot news.

[0018] Figure 2 The present invention is a system diagram of an intelligent collection and management system for hot Internet news.

[0019] Figure 3 The invention is a flow chart of the development trend of the influence analysis coefficient of the intelligent collection and management system of hot news on the Internet. DETAILED DESCRIPTION

[0020] The embodiments of the present invention are described in detail below. The embodiments described below are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention. If no specific techniques or conditions are specified in the embodiments, the techniques or conditions described in the literature in the art or the product specifications are used.

[0021] See also Figure 1 , Figure 2 , an intelligent collection and management system for hot Internet news, including: news collection module, news annotation module, influence analysis module, sentiment analysis module, event prediction module, comprehensive evaluation module and process control module.

[0022] The news collection module captures news information from different news sources for use by subsequent modules. After the news annotation module obtains news information from the news collection module, the news annotation module will perform preliminary processing on each piece of news. The influence analysis module receives data from the news annotation module, and the output results are recorded and passed to the comprehensive evaluation module. The sentiment analysis module also receives news content provided by the news annotation module, and the sentiment analysis results are also recorded and passed to the comprehensive evaluation module. The event prediction module further analyzes the news content, and the prediction results are also passed to the comprehensive evaluation module. The comprehensive evaluation module integrates data from the influence analysis module, sentiment analysis module, and event prediction module; calculates the final evaluation score. Process control module: Determines the management strategy for each piece of news based on the final evaluation score provided by the comprehensive evaluation module.

[0023] The news collection module is used to collect news information from different sources; including locating news sources, automatically crawling or manually entering news content.

[0024] The news annotation module is used to perform preliminary processing on the collected news information, add labels and verify the accuracy of the news information.

[0025] The influence analysis module is used to obtain the number of comments, likes, replies, shares, and readings corresponding to the news information, and evaluate the impact analysis coefficient of the news information; based on the changes in the news information over time, identify the development trend of the news information.

[0026] The sentiment analysis module is used to identify positive, negative and neutral emotions in news, set sentiment scores for the identified emotions, and identify the sentiment impact coefficient corresponding to the news information based on the obtained sentiment scores.

[0027] The event prediction module is used to identify the prediction labels of news information, and classify the news information according to the prediction labels to obtain the comprehensive scores of events corresponding to the news information.

[0028] The comprehensive evaluation module is used to integrate the output results of the influence analysis module, sentiment analysis module and event prediction module to give a final evaluation score; it supports managers to make decisions based on the final evaluation score.

[0029] The process control module is used to obtain the process control parameters corresponding to the news information according to the final evaluation score and adjust the importance and management strategy of the news.

[0030] When collecting news information, the news information from various sources is input into the set collection management system. The input news information needs to include the input time, source, input method, and specific content of the news at the time of input; at this time, news information is automatically captured from different news sources (such as RSS subscriptions, website crawlers, social media APIs, etc.), or manual input of news content is allowed; the collected news information will be stored in a central database or data warehouse for subsequent use.

[0031] The collection methods include but are not limited to real-time interviews, news published by online searches, news manuscripts written by staff, etc., to obtain the news information that needs to be collected at this time.

[0032] In one embodiment of the present invention, when annotating news in the news annotation module, the tags added include geographic location, title theme, picture tag, and text theme to determine the content and specific description method of the current news description; the method for verifying the accuracy of the news information is to verify the current news source, such as verifying the interactive parameters of the publisher, which in this embodiment is the credibility, that is, obtaining the credibility of the publisher, and setting a label for each news information according to the obtained credibility.

[0033] At this time, the calculation method of credibility can be determined based on the ratio of the total number of real news information published by the publisher to the total number of news information published. If the credibility of the news information published by the publisher is higher than the standard credibility, the current news information will be considered credible news. Otherwise, the obtained news information is considered unreliable and there is a certain probability that it is false news information. The selected standard credibility is set to 0.5.

[0034] In one embodiment of the present invention, the influence analysis coefficient in the influence analysis module is obtained by analyzing the reader feedback, number of shares, and reading volume of the current news information; the reader feedback set therein represents the number of comment information corresponding to the current news information, as well as the corresponding number of likes and replies under the corresponding comments, to obtain the reader feedback indicator.

[0035] The impact analysis coefficient can be expressed as: normalize the number of comments, likes, replies, shares, and readings to calculate the impact analysis coefficient.

[0036] I=w C ×C+w L ×L+w R ×R+w S ×S+w V ×V; where I represents the impact analysis coefficient, C represents the number of comments, L represents the number of likes, R represents the number of replies, S represents the number of shares, V represents the number of reads, and w C represents the weight of the number of comments, w L Indicates the weight of the number of likes, w R represents the weight of the number of replies, w S represents the weight of the number of shares, w V Indicates the weight of the reading volume.

[0037] The weights for the number of comments, likes, replies, shares, and readings can be initially set to 0.2, 0.25, 0.15, 0.1, and 0.3, respectively; when calculating the impact analysis coefficient, the selected weights can be modified according to the different news information currently being calculated or the different categories to obtain weights that are more suitable for judging the current news.

[0038] At this point, you can get the impact analysis coefficient of the news when it is disseminated. Based on the obtained impact analysis coefficient, you can know the current situation of news dissemination, so that you can carry out special processing for the identified situations.

[0039] The development trend of news information is to analyze the changing values ​​of the impact analysis coefficient over time and identify the changing values ​​of the impact analysis coefficient in different time periods to quantify the development trend of news.

[0040] The method for identifying the development trend of news information is to form an impact analysis sequence according to the acquired impact analysis coefficients according to the advancement of time points; the method of formation is to convert the impact analysis coefficients into a time series to calculate the development trend of the impact analysis coefficients.

[0041] When calculating the impact analysis sequence, record the time period in the impact analysis sequence that produces an upward trend as the first time period, and the time period in the impact analysis sequence that produces a downward trend as the second time period, determine the number of times the first time period appears and the time period ratio, at this time the time period ratio represents the ratio of each first time period to the set standard time period, to quantify the time period length corresponding to the news information when it is diffused and sent, record the decline rate and the number of occurrences of the second time period, the decline rate of the second time period is determined based on the difference between the maximum and minimum values ​​of the impact analysis coefficient in the second time period; determine the development trend of the impact analysis coefficient based on the first time period and the second time period; output the development trend of the impact analysis coefficient as the development trend of the news information.

[0042] When using the first time period and the second time period to determine the development trend of the impact analysis coefficient, we mainly judge the relative position and probability of the first time period and the second time period to verify the corresponding situations when the news rises and falls at this time. At the same time, we need to verify the range of the value of the change at this time, so as to accurately grasp the time period when the corresponding news loses popularity and the attention decreases.

[0043] like Figure 3 As shown, the implementation method of determining the development trend of the impact analysis coefficient is as follows.

[0044] Firstly, the impact analysis coefficients in the first time period and the second time period are tested using the t-test to obtain the test coefficients.

[0045] Based on the obtained test coefficient, the relative position change between the current news information and other news information in the list is verified when the test coefficient takes different values, and the relative position change rate is obtained.

[0046] The identified inspection coefficients and the corresponding relative position change rates are output as the development trend of the current impact analysis coefficients.

[0047] The format of the above t-test is shown below.

[0048] Among them, t represents the test coefficient, Indicates the difference between the sample data in the first time period and the second time period. At this time, the first time period and the second time period are sampled in equal proportion, and the difference between the impact analysis coefficients in the first time period and the second time period is calculated to determine the stability of the impact analysis coefficient at this time. s d Represents the standard deviation of the sample data, n d Indicates the number of sample data.

[0049] For example, sampling is performed from the currently selected news information, and sampling points are selected in the first time period and the second time period respectively. Then, the impact analysis coefficient can be tested according to the selected points, as shown in Tables 1 and 2 below.

[0050] Table 1 Sampling table

[0051]

[0052] Table 2 Difference table of impact analysis coefficients

[0053] Sampling point The first time period T1 The second time period T2 Difference 1 0.295 0.521 -0.226 2 0.354 0.456 -0.102 3 0.421 0.421 0 4 0.456 0.387 0.069 5 0.489 0.359 0.13 6 0.521 0.33 0.191 7 0.552 0.301 0.251 8 0.584 0.273 0.311

[0054] Then at this time, these selected sampling points will be verified, and the values ​​of the sampling points for testing will be calculated to verify the changes in the impact analysis coefficient at this time. The way to test at this time is to verify by comparing the number of likes. At the same time, it can also be verified according to the increase or decrease in the number of replies to complete the test at this time. Finally, the changes in the five values ​​of the number of comments, likes, replies, shares, and readings will be compared with the preset values ​​in the database to determine the difference between the tested value and the preset value. When the difference is small, it means that the impact analysis coefficient of the test at this time meets the normal setting. If the difference is large, it means that the weight set at this time or the trend of the news being tested is abnormal, and it requires special attention.

[0055] After completing the t-test, the relative position change rate is calculated by calculating the ratio of increase or decrease in the impact analysis coefficient within the fixed time of collection as the relative position change rate at this time; that is, the difference between the impact analysis coefficients of the previous time point and the current time point is taken as the change at this time, and the ratio of the change to the impact analysis coefficient of the previous time point is taken as the output relative position change rate, thereby verifying the change in the corresponding popularity of the current news information as time changes.

[0056] In one embodiment of the present invention, the sentiment analysis module identifies the sentiment impact coefficient in the following manner.

[0057] At this time, we use the sentiment dictionary to conduct sentiment analysis on the acquired news information to obtain the emotions expressed by each piece of news, and set the sentiment impact coefficient for the emotions at this time according to the corresponding data when these emotions are disseminated, so as to judge how this type of news will increase during dissemination and the situation of increase.

[0058] When obtaining the sentiment impact coefficient, the news information is first segmented according to the title and text, and the segmented text is compared with the sentiment dictionary to obtain the sentiment score corresponding to each text. The sentiment polarity score of the current news information is calculated according to the obtained sentiment score. The sentiment polarity score mainly combines words with the same meaning in the news information and assigns scores for positive or negative emotions to these word combinations. All texts in the news information are processed in turn to obtain the score expressed in the entire news, so that subsequent processing can be carried out according to this score.

[0059] At this time, when setting the sentiment score for the text, different scores will be set according to the different emotions expressed by each word. Neutral emotions are set to 0 points, positive emotions are set to 0 to 1 points, and negative emotions are set to -1 to 0 points. At this time, the maximum score of positive emotions is 1, and the maximum score of negative emotions is -1. Neither positive nor negative emotions will take the value of 0, so that the emotions that each text needs to express can be quantified, and the content of the overall news information can be verified based on these scores.

[0060] Therefore, the sentiment polarity score obtained at this time is expressed as the average of the sentiment scores of all texts in the news information, and the number of comments, sharing times, and reading volume are obtained to calculate the propagation coefficient of the news information; the value obtained at this time is a normalized value, and the sentiment influence coefficient is obtained based on the obtained propagation coefficient and sentiment polarity score.

[0061] The sentiment polarity score selected at this time is expressed as the average sentiment score of the entire text, mainly to determine the relative output sentiment in the news. At this time, the sentiment polarity score ranges from -1 to 1, and the set sentiment score also ranges from -1 to 1.

[0062] The propagation coefficient can be expressed as shown below.

[0063] P C =w C ×C+w S ×S+w V ×V; where P C represents the propagation coefficient, C represents the number of comments, S represents the number of shares, V represents the number of readings, and w C represents the weight of the number of comments, w S represents the weight of the number of shares, w V Indicates the weight of the reading volume.

[0064] In this dissemination coefficient, the corresponding number of comments, number of shares and number of readings are the same as those when calculating the impact analysis coefficient. At this time, these three parameters are mainly selected to quantify the corresponding conditions of the corresponding news during dissemination. The weights set at this time will also be adjusted accordingly. For example, according to the order of the number of comments, number of shares and number of readings, the weights can be set to 0.4, 0.3, and 0.3 respectively to quantify the impact of the dissemination at this time.

[0065] The emotional impact coefficient is expressed as shown below.

[0066] Among them, EIC represents the emotional influence coefficient, EPC represents the emotional polarity score, and α represents the adjustment factor, which is used to balance the influence of emotional intensity and communication data. The value range of the selected adjustment factor is 1 to 2. At this time, the adjustment factor is selected based on the value of the emotional polarity score. When the emotional polarity score is in the range of -1 to 0.2, the adjustment factor is 1.2; when the emotional polarity score is in the range of 0.2 to 0.4, the adjustment factor is 1.7; when the emotional polarity score is in the range of 0.4 to 0.5, the adjustment factor is 1.5; when the emotional polarity score is in the range of 0.5 to 0.7, the adjustment factor is 1.3; when the emotional polarity score is in the range of 0.7 to 1.0, the adjustment factor is 1.6. At this time, the selected adjustment factor is mainly affected by the average emotional score of the entire news text when it changes, to quantify how the selected adjustment factor should judge the value at this time.

[0067] Through the calculation at this time, we can know the emotional impact value that the corresponding news may have under the corresponding dissemination, and adjust the management of the news based on this value.

[0068] In one embodiment of the present invention, the event comprehensive score and prediction label set in the event prediction module are used to determine the correlation between news information, such as determining whether the corresponding events in the news information are sub-events of each other, and the classification corresponding to each sub-event. The prediction label is used to determine the classification of the sub-event and the predicted value of subsequent events, so that consecutive or multiple events are managed according to the prediction label, and the predicted classification value corresponding to each sub-event is recorded as the event comprehensive score at this time.

[0069] For example, the labels of the news information at this time are environmental protection policy, new product launch and market feedback. After verifying the existing labels at this time, it is found that environmental protection policy and new product launch are the main event labels. Based on these two main event labels, the sub-events of the current news can be found. At the same time, the time corresponding to the market feedback label is correlated with the time corresponding to other labels. At this time, these correlated labels are the predicted labels at this time, which can be used to judge the development trend of events when events occur in the current news information, and whether there is a correlation between different news; then the correlation between the identified events and sub-events is verified based on the predicted labels, and the correlation value identified at this time is used as the output comprehensive score.

[0070] At this time, the method of identifying the predicted label is to convert the label of the news information into a feature value, and input the feature value into the database as input to determine whether the relevant label corresponding to each feature value exists in the news information. If so, the corresponding label is the predicted label.

[0071] When outputting the comprehensive score of an event, the event prediction module first calculates the correlation parameters between the events corresponding to the news information, and uses the weighted sum of the calculated correlation parameters as the output comprehensive score of the event.

[0072] The comprehensive score for an event is expressed as follows.

[0073] Among them, ECS represents the comprehensive score of the event, β i,j represents the weight of the association parameter between the ith event and the jth event, k i,j Represents the association parameter between the i-th event and the j-th event, i=1,2,,,n-1, j=2,3,,,n, n represents the number of associations between the i-th event and the j-th event.

[0074] The calculation method of the associated parameters at this time is shown as follows.

[0075] Among them, k i represents the eigenvalue corresponding to the i-th event, k j Represents the feature value corresponding to the jth event; at this time, the feature value of the event is determined by extracting the event corresponding to the predicted label in the news information, and setting the feature value according to the type of each event.

[0076] Assuming that a piece of news information predicts a label corresponding to a political event, the following feature values ​​can be set.

[0077] Feature Name: Feature Value Example.

[0078] Event category code: 1 (indicates a political event).

[0079] Location of the incident: "Beijing".

[0080] Event time: 2024-09-2514:00:00 (timestamp).

[0081] Event size: 1000 (number of participants).

[0082] Related Entities: ["Leader A","Government Agency B"].

[0083] Emotional tendency: "positive".

[0084] Media coverage: 500 (number of news reports).

[0085] Precursor event code: None (or a specific precursor event code).

[0086] Time interval between events: None (or a specific time interval).

[0087] Similarity between events: 0.7 (similarity score to another event).

[0088] At this time, after setting the feature value for this time, find the numerical value that can replace the text in these descriptions from the database, and compare these items in turn, so as to obtain the correlation of the events. At this time, the similarity between events is the association parameter calculated above. At this time, the analysis of the corresponding time in the news information is completed, and the comprehensive value of these events is used as the final output event comprehensive score, so as to obtain the final score required. It should be noted that the feature value example given above is only a format for the implementation of the present invention. In actual calculations, the feature name, event category code, event location, event time, event scale, and content corresponding to related entities of the above-set items can be ignored. The calculation is mainly to verify the event interval and the correlation between the events.

[0089] In one embodiment of the present invention, the comprehensive evaluation module normalizes the impact analysis coefficient, the emotional impact coefficient and the comprehensive score of the event for the final evaluation score, and obtains the final evaluation score based on the normalized impact analysis coefficient, the emotional impact coefficient and the comprehensive score of the event.

[0090] Among them, FEC represents the final evaluation score, I represents the impact analysis coefficient, EIC represents the emotional impact coefficient, ECS represents the comprehensive score of the event, λ1 represents the weight of the impact analysis coefficient, λ2 represents the weight of the emotional impact coefficient, λ3 represents the weight of the comprehensive score, and e represents the exponential constant.

[0091] Based on the final evaluation score obtained at this time, combined with the scheme set according to the range of the final evaluation score in the database, it is determined whether the current news needs to be managed and how to adjust the current news management strategy.

[0092] In one embodiment of the present invention, the process control parameters set in the process control module may be release priority, monitoring frequency and feedback response speed.

[0093] Release priority: Based on the news importance score, the release order of news is set to ensure that important news can be presented to the audience first; Monitoring frequency: Different monitoring frequencies are set for news of different importance. Important news requires more intensive monitoring to promptly discover and respond to potential problems. At the same time, the monitoring frequency still includes monitoring the corresponding input news information to promptly follow up on the subsequent content of the news that has occurred; Feedback response speed: According to the feedback from the news audience, the news management team adjusts the processing speed of feedback information to ensure that problems can be solved in a timely manner. In the present invention, the feedback speed is more inclined to adjust the review time and release time of the news to increase the speed of publishing and publicizing the corresponding news.

[0094] At this time, the management strategy output by the process control module is based on the final evaluation score obtained, extracts the process control parameters corresponding to the final evaluation score, compares the process control parameters with the pre-set management strategy, and uses the comparison result as the output management strategy.

[0095] The importance of the news described at this time is determined by the value of the release priority set in the process control parameters; the importance of the output at this time is the release priority corresponding to the management strategy output by the process control module, so as to know the importance of control at this time.

[0096] The method of comparing the process control parameters with the preset management strategies is to use the release priority, monitoring frequency and feedback response speed as the retrieval basis, and obtain the management strategies in the preset management strategies whose similarity with the process control parameters is greater than the preset threshold in turn; the preset thresholds set at this time are set to 0.63, 0.68 and 0.7 in the order of release priority, monitoring frequency and feedback response speed.

[0097] The method for calculating the similarity between the preset management strategy and the process control parameters is to calculate the process control parameters and the standard parameters in the preset management strategy in sequence. The calculation method is to use the probability of selecting the corresponding management strategy when the process control parameters take values ​​to represent the similarity between the process management parameters and the preset management strategy.

[0098] Preset management strategies are a set of processing rules and operation guidelines predefined in the database to guide the release, monitoring and feedback response of news of different importance. These strategies can be formulated based on historical data, industry best practices or specific business needs. Each management strategy usually contains standard parameters in the following aspects: release priority, monitoring frequency, feedback response speed.

[0099] Release priority: Determines the urgency of the news release, such as "immediate release", "regular release", "delayed release", etc.

[0100] Monitoring frequency: Determine the frequency of monitoring news content and its subsequent impact, such as "monitor once an hour", "monitor once a day", "monitor once a week", etc.

[0101] Feedback response speed: determines the response time to user feedback or comments, such as "real-time response", "response within 4 hours", "response within 24 hours", etc.

[0102] Standard parameters are specific indicators in the preset management strategy, which are used to compare with process control parameters.

[0103] For example, standard parameters for release priority might include: immediate release (high priority), regular release (medium priority), delayed release (low priority).

[0104] Standard parameters for monitoring frequency might include: hourly monitoring, daily monitoring, weekly monitoring.

[0105] Standard parameters for feedback response speed may include: real-time response, response within 4 hours, and response within 24 hours.

[0106] The probability of occurrence of a process control parameter refers to the frequency of occurrence of a specific parameter value in the actual news processing process. For example, if the release priority of most news in the system is "regular release", then the probability of occurrence of this parameter value is high. This probability can be obtained by statistically analyzing historical data.

[0107] The probability of occurrence of a standard parameter refers to the frequency with which a certain standard parameter value is set in the preset management strategy. For example, if most preset management strategies set the monitoring frequency to "monitor once a day", then the probability of occurrence of this standard parameter value is high. This probability can also be obtained by statistically analyzing the data of the preset management strategy.

[0108] For example, if the process control parameter is regarded as x and the preset management strategy is regarded as y, then the similarity between the process management parameter and the preset management strategy is as shown below.

[0109] SPS represents the similarity between process management parameters and preset management strategies, p(x,y) represents the probability of co-occurrence of process control parameters and standard parameters in preset management strategies, p(x) represents the probability of occurrence of process control parameters, and p(y) represents the probability of occurrence of standard parameters in preset management strategies.

[0110] At this time, the similarities corresponding to the values ​​of all parameters in the process management parameters are calculated in turn, and the management strategy corresponding to the preset management strategy when the similarity with the process control parameters takes the maximum value is output as the result of comparing the process control parameters with the pre-set management strategy. The comparison result includes the management strategy that needs to be output at this time.

[0111] Although the embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and are not to be construed as limitations on the present invention. A person skilled in the art may make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention and they are still covered by the protection scope of the present invention.

Claims

1. An intelligent collection and management system for hot Internet news, characterized in that: include: News gathering module, used to collect news information from different sources; The news annotation module is used to perform preliminary processing on the collected news information, add labels and verify the accuracy of the news information; The influence analysis module is used to obtain the number of comments, likes, replies, shares, and readings corresponding to the news information, and evaluate the influence analysis coefficient of the news information; according to the changes of the news information over time, the development trend of the news information is identified; Sentiment analysis module, used to identify positive, negative and neutral emotions in news, set sentiment scores for the identified emotions, and identify the sentiment impact coefficient corresponding to the news information based on the obtained sentiment scores; The event prediction module is used to identify the prediction labels of news information, classify the news information according to the prediction labels, and obtain the comprehensive score of the event corresponding to the news information; Comprehensive evaluation module, which is used to integrate the output results of influence analysis module, sentiment analysis module and event prediction module, and give the final evaluation score; The process control module is used to obtain the process control parameters corresponding to the news information according to the final evaluation score and adjust the importance and management strategy of the news.

2. According to claim 1, the intelligent collection and management system for hot Internet news is characterized in that: When the news annotation module annotates news, the tags added include geographic location, title subject, image tag, and text subject.

3. The intelligent collection and management system for hot Internet news according to claim 1 is characterized in that: The impact analysis coefficient can be expressed as follows: normalize the number of comments, likes, replies, shares, and readings to calculate the impact analysis coefficient: I=w C ×C+w L ×L+w R ×R+w S ×S+w V ×V; Among them, I represents the impact analysis coefficient, C represents the number of comments, L represents the number of likes, R represents the number of replies, S represents the number of shares, V represents the number of reads, and w represents the number of comments. C represents the weight of the number of comments, w L Indicates the weight of the number of likes, w R represents the weight of the number of replies, w S represents the weight of the number of shares, w V Indicates the weight of the reading volume.

4. The intelligent collection and management system for hot Internet news according to claim 1 is characterized in that: The way to process the development trend of news information is to form an impact analysis sequence according to the acquired impact analysis coefficients and the progress of the impact analysis coefficients according to the time points; Record the time period that produces an upward trend in the impact analysis sequence as the first time period, and the time period that produces a downward trend in the impact analysis sequence as the second time period, determine the number of times the first time period appears and the time period ratio, record the decline and the number of times the second time period appears, and determine the development trend of the impact analysis coefficient based on the first time period and the second time period; output the development trend of the impact analysis coefficient as the development trend of the news information.

5. The intelligent collection and management system for hot Internet news according to claim 4 is characterized in that: The implementation method of determining the development trend of the impact analysis coefficient is as follows: First, the impact analysis coefficients in the first time period and the second time period are tested using the t-test to obtain the test coefficients; Based on the obtained test coefficient, verify the change in the relative position of the current news information and other news information in the list when the test coefficient takes different values, and obtain the relative position change rate; The identified inspection coefficients and the corresponding relative position change rates are output as the development trend of the current impact analysis coefficients.

6. The intelligent collection and management system for hot Internet news according to claim 1 is characterized in that: The emotional impact coefficient corresponding to the news information is realized as follows: First, the news information is segmented according to the title and text, and the segmented text is compared with the sentiment dictionary to obtain the sentiment score corresponding to each text. The sentiment polarity score of the current news information is calculated according to the obtained sentiment score; The sentiment polarity score is expressed as the average of all text sentiment scores in the news information; Obtain the number of comments, shares, and readings, and calculate the spread coefficient of news information; According to the acquired propagation coefficient and sentiment polarity score, the sentiment influence coefficient is obtained.

7. The intelligent collection and management system for hot Internet news according to claim 1 is characterized in that: The comprehensive score of an event is expressed in the following ways: Calculate the correlation parameters between the events corresponding to the news information, and use the weighted sum of the calculated correlation parameters as the output event comprehensive score; The comprehensive score of the event is expressed as follows: Among them, ECS represents the comprehensive score of the event, β i,j represents the weight of the association parameter between the ith event and the jth event, k i,j represents the association parameter between the ith event and the jth event, i=1,2,,,n-1, j=2,3,,,n, n represents the number of associations between the ith event and the jth event; The calculation method of the associated parameters at this time is expressed as follows: Among them, k i represents the eigenvalue corresponding to the i-th event, k j Represents the eigenvalue corresponding to the jth event.

8. The intelligent collection and management system for hot Internet news according to claim 1 is characterized in that: As for the final evaluation score, the impact analysis coefficient, emotional impact coefficient and event comprehensive score are normalized, and the final evaluation score is obtained according to the normalized impact analysis coefficient, emotional impact coefficient and event comprehensive score: Among them, FEC represents the final evaluation score, I represents the impact analysis coefficient, EIC represents the emotional impact coefficient, ECS represents the comprehensive score of the event, λ1 represents the weight of the impact analysis coefficient, λ2 represents the weight of the emotional impact coefficient, λ3 represents the weight of the comprehensive score, and e represents the exponential constant.

9. The intelligent collection and management system for hot Internet news according to claim 1, characterized in that: The management strategy can be implemented by: Based on the final evaluation score obtained, extract the process control parameters corresponding to the final evaluation score, compare the process control parameters with the pre-set management strategy, and use the comparison result as the output management strategy; The method for comparing the process control parameters with the preset management strategies is to use the release priority, monitoring frequency and feedback response speed as the retrieval basis, and obtain the management strategies in the preset management strategies whose similarity with the process control parameters is greater than the preset threshold in turn; the management strategy corresponding to the maximum similarity with the process control parameters in the preset management strategy is output as the result of comparing the process control parameters with the preset management strategy at this time.

10. The intelligent collection and management system for hot Internet news according to claim 9, characterized in that: The similarities between process management parameters and preset management strategies are shown below: SPS represents the similarity between process management parameters and preset management strategies, p(x,y) represents the probability of co-occurrence of process control parameters and standard parameters in preset management strategies, p(x) represents the probability of occurrence of process control parameters, and p(y) represents the probability of occurrence of standard parameters in preset management strategies.

Images