Public information issuing platform and method

Abstract

The invention discloses a public information publishing platform and method, and relates to the technical field of information public facilities, and the method comprises the following steps: obtaining to-be-evaluated information and publisher data, and analyzing to generate a credibility score PA of an information source; performing sentiment analysis and keyword inspection on the to-be-evaluated information, identifying sentiment deviation and semantic contradiction in the to-be-evaluated information, and analyzing a credibility score PB of the generated content; and acquiring an external authoritative data source, comparing the external authoritative data source with a time consistency check algorithm by using a text similarity algorithm, judging the consistency of the to-be-evaluated information and the external authoritative data source, and outputting a consistency reliability index Pc. According to the method, the problems of low false information identification efficiency and poor accuracy in the prior art are effectively solved by combining multi-dimensional data processing such as information source credibility evaluation, content analysis, external authoritative data source comparison, propagation path analysis and the like.

Description

Technical Field

[0001] The present invention relates to the technical field of public information facilities, and in particular to a public information publishing platform and method. Background Art

[0002] In today's information society, with the widespread popularity of the Internet, the speed of dissemination of various types of information has greatly accelerated. However, the extensiveness and immediacy of information dissemination also bring severe challenges to the authenticity of information. The existence of false information often affects users' decisions and behaviors through large-scale dissemination, seriously damaging the credibility of information platforms and the rights and interests of information consumers.

[0003] To address this problem, traditional methods of false information detection mainly rely on methods such as keyword filtering and rule judgment, but these methods have certain limitations. On the one hand, keyword filtering is easily affected by disguised and variant information, and cannot effectively identify false information that conceals the truth through clever expressions or detail modifications; on the other hand, judgments that rely solely on rules lack in-depth analysis of the complex communication network behind the information, resulting in low efficiency and accuracy in the identification of false information. Therefore, how to accurately and timely identify false information in massive amounts of information, especially through comprehensive judgments from multiple dimensions such as comprehensive information sources, content, and communication paths, has become a core issue that needs to be urgently addressed in the current technical field. Summary of the invention

[0004] The purpose of the present invention is to provide a public information publishing platform and method, which solves the problem that the frequency of false information and rumors spread is increasing year by year, especially in social media, news platforms and various online forums, which not only has a great impact on social order and public trust, but also may trigger a series of political, economic and even social crises. The existence of false information often affects users' decisions and behaviors through large-scale dissemination, seriously damaging the credibility of information platforms and the rights and interests of information consumers.

[0005] The present invention solves the above technical problems through the following technical solutions. The present invention proposes a method for publishing public information, comprising the following steps:

[0006] Step 1: Obtain the information to be evaluated and the publisher data, and analyze and generate the credibility score P of the information source A ;

[0007] Step 2: Use sentiment analysis and keyword verification on the information to be evaluated to identify sentiment deviations and semantic contradictions in the information to be evaluated, and analyze the credibility score P of the generated content. B ;

[0008] Step 3: Obtain external authoritative data sources, and use text similarity algorithms and time consistency checking algorithms for comparison to determine the consistency between the information to be evaluated and the external authoritative data sources, and output the consistency reliability index P c ;

[0009] Step 4: Analyze the information dissemination path, and use social network analysis techniques to detect information dissemination patterns and abnormal user behaviors, and output the reliability index P of the dissemination path D ;

[0010] Step 5: Based on the credibility score P of the information source A , the credibility score P of the content B , the consistency reliability index P c , and the reliability index P of the dissemination path D , use a comprehensive credibility scoring model to comprehensively evaluate the credibility of the information to be evaluated, and obtain a comprehensive credibility score. If the comprehensive credibility score is lower than the set credibility threshold, it is determined as false information;

[0011] Step 6: According to the judgment result, take processing measures for the information to be evaluated

[0012] Preferably, the generation steps of the credibility score P of the information source A include:

[0013] S101: Collect data of the information to be evaluated: Obtain the specific content and publisher data of the information to be evaluated from the information release platform. The publisher data includes the historical behavior data and interaction data of the publisher;

[0014] S102: Selection and quantification of core parameters: Select the core parameters of the publisher as the basis for calculating the credibility score P of the information source A . The core parameters of the publisher include:

[0015] Publisher historical trustworthiness P1: Based on the verification results of the information published by the publisher in the past, calculate its credibility. The specific formula is as follows:

[0016]

[0017] where i is the sequential number of historical information, V i is the credibility score of each historical information, and n is the number of historical information;

[0018] Information source category score P2: Assign weights according to the information source type, and obtain the information source category score of each information source type through historical analysis;

[0019] Interaction feedback index P3: By analyzing the interaction feedback of users after the information to be evaluated is published, generate the interaction feedback index P3. The specific formula is as follows:

[0020]

[0021] Among them, C is the number of comments, L is the number of likes, S is the number of shares, and E is the entropy of the interaction type distribution;

[0022] The entropy E of the interaction type distribution is:

[0023] E = -(P C lnP C + P L lnP L + P S lnP S )

[0024] Among them, are respectively the proportions of the number of comments, the number of likes, and the number of shares in the total number of interactions;

[0025] T is the total number of interactions, and it is: T = C + L + S;

[0026] S103: Construct a credibility scoring model for the information source: Integrate the publisher's historical trust degree P1, the information source category score P2, and the interaction feedback index P3 obtained in S102, and calculate the credibility score P of the information source through the following formula A as:

[0027] P A = w1·P1 + w2·P2 + w3·P3

[0028] Among them, w1, w2, and w3 are the weight coefficients of the publisher's historical trust degree P1, the information source category score P2, and the interaction feedback index P3.

[0029] Preferably, the steps for generating the credibility score P of the content B include:

[0030] S201: Conduct sentiment analysis on the content in the information to be evaluated, specifically as follows:

[0031] The sentiment analysis of the information to be evaluated outputs the sentiment category and sentiment intensity; and by comparing the sentiment distribution of the content of the information to be evaluated with the conventional pattern, it detects whether there is a sentiment deviation; and gives the corresponding sentiment analysis score P4, and its calculation formula is:

[0032] P4 = sign(H)·|H|

[0033] Among them, H is the sentiment intensity, and sign(H) is the polarity of the sentiment;

[0034] S202: Extract the key information from the information to be evaluated, identify the negative keywords in the information to be evaluated, and mark them as potential false information. The calculation formula for the keyword inspection score P5 is:

[0035]

[0036] where K q is the q-th keyword, m is the total number of all keywords, Weight(K q ) is the weight of the keyword in false information, and Frequency(K q ) is the frequency of the keyword in the text;

[0037] S203: If semantic contradictions are detected during the analysis process, mark them as semantically inconsistent and reduce the semantic consistency score. The calculation formula for the semantic consistency score P6 is:

[0038]

[0039] where Conflict-Score is the intensity of semantic contradictions and N is the number of contradictions in the information to be evaluated;

[0040] S204: Calculate the credibility score P B of the content. The calculation formula for the credibility score P B of the content is:

[0041] P B = w4·P4 + w5·P5 + w6·P6

[0042] where w4, w5, and w6 are the weight coefficients of the sentiment analysis score P4, the keyword inspection score P5, and the semantic consistency score P6.

[0043] Preferably, the specific steps of step three are:

[0044] S301: Obtain data related to the content of the information to be evaluated from multiple external authoritative data sources;

[0045] S302: The text similarity algorithm is used to compare the information to be evaluated with the information from external authoritative data sources, and the text similarity algorithm is used to measure the content consistency between the information to be evaluated and the information from external authoritative data sources. The specific steps are as follows:

[0046] First, preprocess the information to be evaluated and the external authoritative data sources;

[0047] Subsequently, use the cosine similarity CS to compare the information to be evaluated with the information from external authoritative data sources. Assume that the text of the information to be evaluated is T1 and the text of the external authoritative data source is T2. Then the formula for the text similarity CS(T1, T2) is:

[0048]

[0049] Among them, T1,j and T2,j are the weights or word frequencies of the j-th word in texts T1 and T2 respectively, and h is the number of different features in the information to be evaluated and the text of the external authoritative data source;

[0050] S303: For the information to be evaluated containing time information, adopt the time consistency test algorithm to ensure that the time of information release is consistent with the time of event occurrence, including the following processes:

[0051] Extract the key time data from the information to be evaluated;

[0052] Compare the time data of the information to be evaluated with the corresponding event occurrence time in the external authoritative data source, and the probability P of the authenticity of the time data of the information to be evaluated time is:

[0053]

[0054] Among them, ΔT = |T eval -T ref | is the time difference between the information to be evaluated and the external authoritative data source, and T eval is the release time of the information to be evaluated, and T ref is the corresponding event occurrence time of the external authoritative data source; μ is the mean of the time difference between the information to be evaluated and the external authoritative data source, σ is the standard deviation of the time difference, and T t is the threshold of the time difference;

[0055] S304: Synthesize the text similarity CS(T1, T2) and the probability P of the authenticity of the time data time , and generate the consistency reliability index P c :

[0056] P c = CS(T1, T2)·P time

[0057] If the consistency reliability index P c meets the requirements of the consistency reliability index threshold, the information to be evaluated is considered to be consistent with the external authoritative data source and has a high credibility; if there is a large inconsistency, the information to be evaluated is false information or needs to be further verified.

[0058] Preferably, the method for setting the threshold of the consistency reliability index is: set the consistency reliability index threshold T threshold :

[0059] Tthreshold = μ + k·σ

[0060] Where k is the misjudgment constant.

[0061] Preferably, step four includes:

[0062] S401: Obtain the propagation data related to the information to be evaluated from the social network platform and construct a propagation network through user interaction data;

[0063] S402: Detect the propagation mode: Based on the constructed propagation network, use social network analysis algorithms to detect the propagation mode of the information and calculate the propagation centrality of each node to identify the key nodes in the information propagation; The calculation formula for the node propagation centrality is:

[0064]

[0065] Where C(v) is the propagation centrality of node v, d(u, v) is the propagation distance from node u to node v, u ∈ V is the set of nodes to which node u belongs in the entire social network, and V is the set of all nodes in the social network;

[0066] S403: Detect abnormal user behavior:

[0067] User behavior analysis: Based on the interaction data of users on the social platform, use behavior analysis techniques to detect whether there are abnormal user behaviors. The specific calculation formula for the abnormal behavior score BAS(v) is:

[0068]

[0069] Where B s is the behavior characteristic of the s-th user, is the average value of user behaviors, N is the total number of users. If the score exceeds the set abnormal behavior score threshold, the behavior is determined to be abnormal;

[0070] Influence analysis in the social network: Through influence analysis, identify the key influencers in the propagation of the information to be evaluated. The calculation formula for the node influence I(v) is:

[0071]

[0072] Where N (v) is the set of neighbor nodes of node v, d(u, v) represents the propagation distance between node u and node v, and S(u) is the propagation ability of node u;

[0073] S404: The propagation path reliability index P D The calculation formula is:

[0074]

[0075] Among them, GNN is a graph neural network used to learn the propagation path features between nodes;

[0076] Self-Attention is a self-attention mechanism used to process time consistency checks, evaluate the consistency between time differences and external data sources, and combine with P time to perform information reliability correction.

[0077] Preferably, the abnormal behavior scoring threshold is dynamically adjusted according to the real-time data changes of user behavior, and the calculation formula of the abnormal behavior scoring threshold is:

[0078] T abnormal (t) = T abnormal (t - 1) + α·ΔB(t)

[0079] where, T abnormal (t) is the threshold at time t, ΔB(t) is the increment according to the change of user behavior, and α is the learning rate.

[0080] Preferably, the steps for the comprehensive credibility scoring model to comprehensively evaluate the information credibility are as follows:

[0081] S501: Establish a comprehensive credibility scoring model and calculate the comprehensive credibility score P 终 :

[0082] P 终 = w7·P A + w8·P B + w9·P C + w 10 ·P D

[0083] where, w7, w8, w9, w 10 are the weight coefficients of P A , P B , P C , P D respectively;

[0084] S502: Set the credibility threshold. When the comprehensive credibility score is higher than the credibility threshold, the information is considered credible; when the comprehensive credibility score is lower than the credibility threshold, the information is considered likely to be false information;

[0085] S503: The result of the comprehensive credibility score will provide a basis for subsequent false information disposal. False information will be subject to corresponding restrictive measures, and credible information will continue to be published.

[0086] Preferably, the comprehensive credibility score is dynamically optimized: by continuously collecting the feedback of platform users, the dissemination effect of information, and the recognition results, the comprehensive credibility scoring model is regularly optimized, and through machine learning and data mining technologies, the weight coefficients of P A 、P B 、P C 、P D are adjusted.

[0087] The present invention also proposes a platform for a public information release method, including:

[0088] An information collection module, which is used to obtain the information to be evaluated and the publisher data, and collect the historical behavior records, interaction data, and information source category scores of the information publisher;

[0089] A content analysis module, which is used to perform sentiment analysis on the information to be evaluated, and analyze the sentiment deviation and semantic contradiction in the text;

[0090] An external authoritative data source comparison module, which is used to obtain an external authoritative data source, and use a text similarity algorithm to compare the information content to check the consistency of the content with the external authoritative data source;

[0091] A propagation path analysis module, which is used to use social network analysis technology to construct an information propagation network, analyze the propagation path and propagation speed of the information, and detect whether there is an abnormal information propagation mode;

[0092] A comprehensive credibility scoring module, which is used to comprehensively evaluate the credibility of the information by adopting a comprehensive credibility scoring model for the credibility scores P A 、P B 、P c of the content, the consistency reliability index P D and the propagation path reliability index P

[0093] An information processing module, which executes corresponding information processing measures according to the false information determination result output by the comprehensive credibility scoring module.

[0094] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0095] The present invention effectively solves the problems of low false information recognition efficiency and poor accuracy in the prior art by combining multi-dimensional data processing such as information source credibility evaluation, content analysis, comparison with external authoritative data sources, and dissemination path analysis. Through the comprehensive evaluation of the historical behavior, interaction records, and information source category scores of the information source, the credibility basis of information release is ensured; the authenticity of the content is deeply mined by combining sentiment analysis and keyword verification; the comparison with external authoritative data sources eliminates the false risks brought by the inconsistency between the information and external authoritative data sources through text similarity and time consistency tests; the social network analysis technology effectively detects the dissemination path of information and abnormal user behavior, revealing the abnormal information dissemination mode of false information. Finally, through the comprehensive evaluation of the comprehensive credibility scoring model, the credibility of the information is accurately judged, and false information processing measures are automatically executed, reducing manual intervention and improving the efficiency and accuracy of information management. Compared with the traditional rule-based or keyword filtering methods, the present invention can dynamically adapt to the diversity of information content and dissemination, greatly improving the recognition accuracy and processing speed of false information, and ensuring the reliability of platform content and user trust. BRIEF DESCRIPTION OF THE DRAWINGS

[0096] Figure 1 It is a schematic flowchart of the public information release method of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0097] The following further describes the above and other technical features and advantages of the present invention in more detail with reference to the accompanying drawings.

[0098] This embodiment provides a technical solution: a public information release method, as Figure 1 shown, including the following steps:

[0099] Step 1: Obtain the information to be evaluated and the publisher data, and analyze and generate the credibility score P of the information source A ; the specific steps are as follows:

[0100] S101: Collect the data of the information to be evaluated: Obtain the specific content of the information to be evaluated and the publisher data from the information release platform; the publisher data includes, but is not limited to, the historical behavior data and interaction data of the publisher (such as comments, likes, shares, etc.), historical trust scores, the types of content published (such as news, advertisements, social information, etc.), and information source category scores (such as official media, individual users, third-party platforms, etc.);

[0101] By collecting detailed data on the information to be evaluated and its publisher, a comprehensive understanding of the information's release background, the publisher's historical behavior, and post-release user interactions can be achieved. This step provides comprehensive data support for subsequent comprehensive credibility scoring, ensuring that the evaluation of information sources does not rely solely on a single dimension but comprehensively evaluates the reliability of publishers through multiple dimensions (such as historical behavior, interaction feedback, etc.). This helps identify the publisher's behavior patterns, further distinguish the credibility of information sources, and avoid biases resulting from solely relying on publisher identification or evaluation criteria for individual information.

[0102] S102: Selection and Quantification of Core Parameters: Select the core parameters of the publisher as the basis for calculating the credibility score P of the information source. The core parameters of the publisher include: A of the information source. The core parameters of the publisher include:

[0103] Publisher's Historical Trustworthiness P1: Calculate its credibility based on the verification results of the information published by the publisher in the past. The specific formula is as follows:

[0104]

[0105] where i is the sequential number of historical information, V i is the credibility score of each piece of historical information, and n is the number of historical information;

[0106] Information Source Category Score P2: Assign weights according to the information source type (such as official media, well-known websites, etc.), and obtain the information source category score for each information source type through historical analysis. For example:

[0107]

[0108] Interaction Feedback Index P3: Generate the interaction feedback index P3 by analyzing the user interaction feedback (such as comments, likes, shares, etc.) after the information to be evaluated is released, which reflects the influence of the information on the audience. Its calculation method is weighted average, that is:

[0109]

[0110] where C is the number of comments, L is the number of likes, S is the number of shares, and E is the entropy of the interaction type distribution;

[0111] The entropy E of the interaction type distribution is:

[0112] The entropy E of the interaction type distribution is:

[0113] E = -(P C lnP C + P L lnP L + P S lnP S )

[0114] Among them, They are the proportions of the number of comments, likes, and shares to the total number of interactions respectively; T is the total number of interactions, and it is: T = C + L + S;

[0115] By selecting and quantifying core parameters (such as the historical trustworthiness of the publisher, the information source category score, and the interaction feedback index), the credibility score of the information source can be accurately calculated in a quantitative manner. The historical trustworthiness of the publisher verifies the historical published content of the publisher and the authenticity of the information to ensure that the credibility score is based on facts and data. The information source category score helps to establish the association between the information source type and authority, thus providing a quantifiable classification standard for scoring. The interaction feedback index reflects the user's acceptance degree and feedback quality of the information. Through this quantitative means, it can effectively evaluate the influence of the information source on users and its dissemination effectiveness, so as to ensure that the subsequent identification of false information can rely on multiple credible dimensions.

[0116] S103: Construct a credibility scoring model for the information source: Integrate the parameters obtained in S102 (i.e., the historical trustworthiness of the publisher P1, the information source category score P2, and the interaction feedback index P3), and calculate the credibility score P of the information source through the following formula A It is:

[0117] P A = w1·P1 + w2·P2 + w3·P3

[0118] Among them, w1, w2, and w3 are the weight coefficients of the historical trustworthiness of the publisher P1, the information source category score P2, and the interaction feedback index P3 respectively. The specific values of the weight coefficients can be dynamically adjusted according to historical data and user feedback.

[0119] The setting of the weight coefficients is optimized through historical data and user feedback to ensure that the credibility scoring model of the information source fits the actual application scenario and reflects the performance of different types of information sources in various information releases. Through the application of the credibility scoring model of the information source, the data of each dimension can be more accurately integrated to form a reliable information source score, providing an evaluation result of credibility, and further laying a data foundation for the authenticity judgment of information and the identification of false information.

[0120] Step 2: Use sentiment analysis and keyword checking for the information to be evaluated, identify the sentiment deviation and semantic contradiction in the information to be evaluated, and analyze and generate the credibility score P of the content B ;

[0121] The generation steps of the credibility score P of the content B are as follows:

[0122] S201: Conduct sentiment analysis on the content in the information to be evaluated. Through natural language processing (NLP) techniques, based on deep learning models (such as BERT, LSTM), extract the sentiment features in the information to be evaluated. The goal of sentiment analysis is to identify the potential sentiment polarity (positive, negative, or neutral) in the text. Specifically as follows:

[0123] Classify the information to be evaluated, including sentiment categories and sentiment intensity (e.g., the positive and negative scores of the text); and detect whether there is a sentiment deviation by comparing the sentiment distribution of the content of the information to be evaluated with the normal pattern; for example, if certain texts contain words with extreme emotions or expressions of emotional conflicts (such as "I support this, but I hate it"), it is determined as sentiment inconsistency; and give the corresponding sentiment analysis score P4, and its calculation formula is:

[0124] P4 = sign(H)·|H|

[0125] Where H is the sentiment intensity, and sign(H) is the polarity of the sentiment (positive, negative, or neutral). If the sentiment intensity is too high and tends to be negative or overly radical, a low score is given;

[0126] By identifying the sentiment polarity (positive, negative, neutral) of the text, it is possible to effectively distinguish normal neutral information from emotional information. If the text sentiment is overly biased (such as using words with extreme emotions), it may indicate that the information has potential misleading or false components. Through sentiment analysis, it is possible to provide the transparency of the sentiment for the information, making it easier to be recognized and verified, and providing an effective basis for the subsequent credibility assessment of the information.

[0127] S202: Extract the key information in the information to be evaluated and identify possible semantic contradictions or sentiment deviations; First, identify the key entities (such as people, places, events, etc.) in the text through named entity recognition (NER) technology. Then, construct a specific keyword library, including keywords reflecting different emotions and semantic contradictions (e.g.: "scam", "false", "misleading", etc.). If a negative keyword appears in the information to be evaluated, it is further marked as potentially false information, and the calculation formula for the keyword inspection score P5 is:

[0128]

[0129] Where K q is the qth keyword, m is the total number of all keywords, Weight(K q ) is the weight of the keyword appearing in false information, and Frequency(K q ) is the frequency of the keyword appearing in the text;

[0130] Extracting key entities (such as people, places, events, etc.) in the information to be evaluated through Named Entity Recognition (NER) technology can help identify potential false components in the information, especially to check whether there are self - contradictions or semantic inconsistencies in the information. Semantic contradictions may be one of the common features of false information. Therefore, by identifying and marking these contradictions, information bias and misleading effects can be reduced. If the text contains expressions that conflict with the conventional semantic patterns (such as contradictory statements in different paragraphs), this analysis process can effectively help reduce the credibility score of the information and further screen potential false information.

[0131] S203: If semantic contradictions are detected during the analysis process (for example, two opposite expressions like "support this policy" and "this policy is not applicable" appear in the same information), mark it as semantically inconsistent and reduce the semantic consistency score. The calculation formula for the semantic consistency score P6 is:

[0132]

[0133] Among them, Conflict - Score is the intensity of semantic contradictions, N is the number of contradictions appearing in the information to be evaluated. If there are more contradictions, the final score is close to 0, indicating that the credibility of the information is low.

[0134] By detecting semantic contradictions and marking them as semantically inconsistent, it is possible to further accurately screen out forged or misleading content in the information. Semantic consistency analysis ensures the logic and consistency within the information, which is crucial for identifying false information. If there are internal semantic contradictions in the information, it may indicate that the information has been tampered with or is inaccurate, thus reducing its credibility. This processing link plays a key role in information analysis and can provide accurate data for subsequent false information identification through text consistency, providing important support for the authenticity judgment of information.

[0135] S204: Calculate the credibility score P of the content B , the credibility score P of the content B The calculation formula is:

[0136] P B = w4·P4 + w5·P5 + w6·P6

[0137] Among them, w4, w5, w6 are the weight systems of the sentiment analysis score P4, the keyword test score P5, and the semantic consistency score P6. If the credibility score P of the content B is low, it indicates that there are strong sentiment biases, semantic contradictions or false - related keywords in the information, and it needs to be further marked as suspicious or false.

[0138] The detection accuracy of information authenticity is greatly improved through multi-dimensional anomaly analysis (including sentiment deviation and semantic contradiction), ensuring the content quality of the information platform and providing effective support for false information filtering and risk management of the system.

[0139] Step 3: Obtain external authoritative data sources, use text similarity algorithms and time consistency verification algorithms to compare the information, judge the consistency between the information to be evaluated and the external authoritative data sources, and output the consistency reliability index P c ;

[0140] The specific steps of Step 3 are as follows:

[0141] S301: Obtain data related to the content of the information to be evaluated from multiple external authoritative data sources (such as government websites, news media, academic databases, authoritative announcements, etc.). These external authoritative data sources must be verified public credibility institutions or platforms to ensure the authenticity and authority of their content. The information of external authoritative data sources can include news reports, historical event data, government announcements, expert evaluations, etc. The main purpose is to ensure a high degree of consistency between the information to be evaluated and external authoritative sources;

[0142] S302: The text similarity algorithm is used to compare the information to be evaluated with the information of the external authoritative data source, and text similarity algorithms (such as cosine similarity, Jaccard similarity, etc.) are used to measure the content consistency between the information to be evaluated and the information of the external authoritative data source. The specific steps are as follows:

[0143] First, preprocess the information to be evaluated and the external authoritative data source, including operations such as removing stop words, word segmentation, and stemming;

[0144] Subsequently, use cosine similarity CS to compare the information to be evaluated with the information of the external authoritative data source. Assume that the text of the information to be evaluated is T1 and the text of the external authoritative data source is T2. Then the formula for text similarity CS(T1, T2) is:

[0145]

[0146] Among them, T1,j and T2,j are the weights or word frequencies of the jth word in texts T1 and T2 respectively, and h is the number of different features in the texts of the information to be evaluated and the external authoritative data source. This formula calculates the cosine value of the angle between two text vectors, and the score range is from 0 to 1. The closer the value is to 1, the more consistent the text content;

[0147] If the similarity is higher than the set threshold (for example, 0.85), it is determined that the information is consistent; if the similarity is low, the information is further marked as suspicious;

[0148] S303: For the information to be evaluated that contains time information, use the time consistency check algorithm to ensure that the time of information release is consistent with the time of event occurrence, including the following processes:

[0149] Extract key time data (such as event occurrence time, news release date, etc.) from the information to be evaluated;

[0150] Compare the time data of the information to be evaluated with the corresponding event occurrence time in the external authoritative data source. For example, if the release time of a certain news event is January 1, 2025, and the actual occurrence time of the event is December 25, 2024, then there is a risk of time inconsistency in the information;

[0151] Use time difference calculation. If the time difference exceeds the preset threshold (for example, 1 hour, 1 day, etc.), then mark it as information inconsistent, and the probability P of the authenticity of the time data of the information to be evaluated time is:

[0152]

[0153] where, ΔT = |T eval -T ref | is the time difference between the information to be evaluated and the external authoritative data source, and T eval is the release time of the information to be evaluated, T ref is the corresponding event occurrence time of the external authoritative data source; μ is the mean of the time difference between the information to be evaluated and the external authoritative data source, generally 0, σ is the standard deviation of the time difference, representing the range of time data fluctuations, T t is the threshold of the time difference, indicating the maximum allowable time difference, and the time difference exceeding this value will be regarded as untrustworthy;

[0154] Compare the time data in the information to be evaluated with the corresponding time in the external authoritative data source, and use the time consistency check algorithm (such as time difference calculation) to check the consistency between the information release time and the actual event occurrence time. If the time difference exceeds the preset threshold, it may indicate that there is an error or false component in the information content.

[0155] S304: Generate the consistency reliability index P based on the results of text similarity and time consistency check c :

[0156] P c = CS(T1, T2)·P time

[0157] If the consistency reliability index P cIf the consistency reliability index threshold requirement is met, the information to be evaluated is considered consistent with the external authoritative data source and has a relatively high credibility. If there are significant inconsistencies, the information to be evaluated is false information or needs further verification.

[0158] The method for setting the threshold of the consistency reliability index is as follows: Set the consistency reliability index threshold T by using the mean μ of the time difference between the information to be evaluated and the external authoritative data source based on the normal distribution and the standard deviation σ of the time difference. threshold :

[0159] T threshold = μ + k·σ

[0160] Among them, k is the misjudgment constant (such as 1 or 2), which controls the misjudgment tolerance. This method is applicable to scenarios with sufficient data and can automatically adjust the consistency reliability index threshold according to historical information.

[0161] Use the consistency reliability index P c as the basis for subsequent judgment of information credibility. If the consistency reliability index P c is of low consistency, the system will further mark this information as potentially false information and enter the subsequent steps for further processing.

[0162] By obtaining the external authoritative data source and comparing it with the information to be evaluated, and using the text similarity algorithm and the time consistency test algorithm to verify the authenticity of the information, the ability to identify false information is effectively enhanced. The acquisition of the external authoritative data source provides a credible reference for the information, making the comparison more objective and fair, and avoiding the deviation that may be brought by relying on a single information source. The text similarity comparison uses the algorithm to effectively measure the matching degree between the information to be evaluated and the authoritative data source to ensure the consistency of the information content, while the time consistency test ensures that the time of information release is consistent with the actual event, reducing the misleading caused by false information. Through the comprehensive application of these two algorithms, it is possible to accurately identify whether the information is consistent with the external authoritative data. If the consistency is low, it indicates that there are certain false components or misleading content in the information.

[0163] Step Four: Analyze the information dissemination path, use social network analysis technology to detect the information dissemination pattern and abnormal user behavior, and output the dissemination path reliability index P D ;

[0164] The specific steps of Step Four are as follows:

[0165] S401: Obtain the dissemination data related to the information to be evaluated from social network platforms (such as Weibo, Twitter, Facebook, etc.) and construct a dissemination network through user interaction data; the dissemination data includes records of interactive behaviors such as information sharing, forwarding, commenting, and liking, and particularly focuses on the dissemination path, timestamp, and user participation of the information; constructing a dissemination network through user interaction data is as follows: The behaviors of each user on the social platform (such as forwarding, commenting, liking) can be regarded as a node in the dissemination chain. The information starts from the publisher and is disseminated through other users in sequence, forming a directed graph. The nodes represent users, and the edges represent the information dissemination between users. The construction formula of the dissemination network diagram is:

[0166] G=(V,E)

[0167] Among them, V is the set of nodes (i.e., platform users), E is the set of edges (i.e., the information dissemination path between users). By calculating the dissemination centrality of the nodes, evaluate the efficiency and suspiciousness of information dissemination, and identify the core nodes of information dissemination and their influence;

[0168] S402: Detect the dissemination pattern: Based on the constructed dissemination network, use social network analysis algorithms (such as PageRank, HITS, dissemination tree analysis, etc.) to detect the dissemination pattern of information. Usually, the dissemination of false information shows different rules from normal information, such as the nodes of information dissemination being more concentrated and the dissemination path being shorter, etc. By analyzing indicators such as the speed, scope, and centrality of information dissemination, judge the abnormality of dissemination; and calculate the dissemination centrality of each node to identify the key nodes in information dissemination (i.e., the users with the greatest influence); Nodes with higher centrality are often the sources of false information dissemination. The calculation formula of dissemination centrality is:

[0169]

[0170] Among them, C(v) is the dissemination centrality of node v, d(u, v) is the dissemination distance from node u to node v, u∈V is that node u belongs to the set of nodes of the entire social network, V is the set of all nodes in the social network. If the calculated value of centrality is abnormally high, it indicates that information dissemination is concentrated and there may be human intervention;

[0171] In addition, through the analysis of the dissemination pattern, identify whether there are abnormal dissemination behaviors of information, such as the dissemination path being too concentrated, and abnormal dissemination behaviors of certain specific users or nodes (for example, a large number of forwards or comments in a short period of time). This abnormal pattern usually indicates that the information is maliciously manipulated or intervened by robot accounts;

[0172] S403: Detect abnormal user behaviors:

[0173] User Behavior Analysis: Based on the interaction data of users on social platforms (such as comments, shares, likes, etc.), behavioral analysis techniques (such as clustering analysis, anomaly detection algorithms) are used to detect whether there are abnormal user behaviors. For example, behaviors such as a large number of repeated forwards of the same content in a short period of time, and a high consistency between likes and comments may indicate that the information is manipulated by machine algorithms or there is malicious human operation. The specific calculation formula for the abnormal behavior score BAS(v) is:

[0174]

[0175] where B s is the behavior characteristic of the s-th user (such as forwarding frequency, number of comments, etc.), is the average value of user behaviors, N is the total number of users. If the score exceeds the set abnormal behavior score threshold, the behavior is determined to be abnormal;

[0176] Influence Analysis in Social Networks: Through influence analysis, the key influencers in the spread of the information to be evaluated are identified. Nodes with greater influence are usually the sources of information dissemination. False information often spreads rapidly through these highly influential users. The calculation formula for the node influence I(v) is:

[0177]

[0178] where is the set of neighbor nodes of node v, d(u, v) represents the propagation distance between node u and node v, and S(u) is the propagation ability (spread degree) of node u, which can be measured according to the connectivity of node u or historical propagation situations (such as behaviors like forwards, likes, etc.). The propagation ability S(u) can be approximately represented by the degree of the node (i.e., the number of other nodes directly connected to this node), or comprehensively calculated by combining data such as interaction behaviors and influence in the social network;

[0179] S404: Propagation Path Reliability Index P D The calculation formula is:

[0180]

[0181] where GNN is a graph neural network used to learn the propagation path characteristics between nodes. It dynamically evaluates the reliability of the information propagation path based on multi-dimensional information such as the propagation centrality C(v), abnormal behavior score BAS(v), and node influence I(v) of the nodes;

[0182] Self-Attention is a self-attention mechanism used to handle time consistency verification, evaluate the consistency between time differences and external data sources, and combine P time to perform information reliability correction.

[0183] The abnormal behavior scoring threshold is dynamically adjusted according to the real-time data of user behavior. For example, using incremental learning algorithms (such as online learning algorithms), the threshold is automatically adjusted each time new data is input. Whenever a large number of users are found to exhibit abnormal behavior, the system will automatically lower the threshold to make it more sensitive; in normal situations, the system will automatically increase the threshold to reduce false alarms. The advantages are as follows: dynamically adapting to different data patterns, ensuring that the system can respond to changes in user behavior in real time, and reducing human intervention; the calculation formula for the abnormal behavior scoring threshold is:

[0184] T abnormal (t) = T abnormal (t - 1)+α·ΔB(t)

[0185] where, T abnormal (t) is the threshold at time t, ΔB(t) is the increment according to the change in user behavior, and α is the learning rate.

[0186] Step four provides a comprehensive analysis of the propagation path and behavior anomalies, which can accurately detect abnormal features in information dissemination and identify whether false information is spread through unnatural channels. This analysis helps the platform discover potential false information sources in the early stage of information dissemination and provides solid data support for subsequent false information evaluation and processing, ensuring the credibility and accuracy of the platform's information.

[0187] Step five: Based on the relevant data obtained in steps one to four (the credibility score P A of the information source, the credibility score P B of the content, the consistency reliability index P c of the content, the propagation path reliability index P D ), a comprehensive credibility scoring model is used to comprehensively evaluate the credibility of the information, and a comprehensive credibility score is obtained. If the comprehensive credibility score is lower than the set credibility threshold, it is determined as false information;

[0188] In steps one to four, each module provides data in multiple dimensions (the credibility score P A of the information source, the credibility score P B of the content, the consistency reliability index P c of the content, the propagation path reliability index P D ). By inputting this data in different dimensions into the comprehensive credibility scoring model, various information sources and analysis results can be effectively integrated, avoiding biases or deficiencies in a single dimension, and ensuring a more comprehensive and objective evaluation of information credibility. Through the comprehensive credibility scoring model, the weights of different factors can be flexibly adjusted, enabling certain important dimensions to have a greater impact on the final score, thereby improving the accuracy of the evaluation.

[0189] The steps for comprehensively evaluating the credibility of information using a comprehensive credibility scoring model are as follows:

[0190] S501: Based on the relevant data obtained in the previous steps (the credibility score P of the information source A , the credibility score P of the content B , the consistency reliability index P c , the reliability index P of the dissemination path D ), establish a comprehensive credibility scoring model to comprehensively consider the influence of each index. The comprehensive credibility scoring model will consider the importance of each piece of data to the final information credibility and calculate the comprehensive credibility score P 终 :

[0191] P 终 = w7·P A + w8·P B + w9·P C + w 10 ·P D

[0192] where w7, w8, w9, w 10 are the weight coefficients of P A , P B , P C , P D respectively, and w 10 , w 11 , w 12 , w 13 are dynamically optimized according to historical data and user feedback to reflect the actual importance of each relevant data in practical applications. For example, the credibility score P of the content B may be more important than the dissemination path analysis in some scenarios, so the corresponding weight will be larger;

[0193] S502: Set a credibility threshold. When the comprehensive credibility score is higher than the credibility threshold, the information is considered credible; when the comprehensive credibility score is lower than the credibility threshold, the information may be considered false information;

[0194] By continuously optimizing the model, the credibility threshold can be dynamically adjusted according to the actual situation. For example, if the false rate of a certain type of information is relatively high, the credibility threshold can be lowered to improve the recognition rate of false information; conversely, the credibility threshold can be raised to reduce misjudgment.

[0195] S503: The result of the comprehensive credibility score will provide a basis for subsequent handling of false information. Corresponding restrictive measures (such as automatic deletion, restricted dissemination, or warning the publisher, etc.) will be taken for false information, and credible information will continue to be published.

[0196] Dynamic Optimization of Comprehensive Credibility Score: By continuously collecting feedback from platform users, information dissemination effects, and recognition results, regularly optimize the comprehensive credibility scoring model. Through machine learning and data mining techniques, adjust the weights of various scores so that the comprehensive credibility scoring model can better adapt to different types of information and the changing characteristics of false information.

[0197] The application of the comprehensive credibility scoring model greatly enhances the ability to identify false information. By objectively combining data from different dimensions, the credibility assessment of information not only considers the reliability of the information source but also takes into account the emotional bias of the content, semantic contradictions, the matching degree with external authoritative data sources, and abnormal behaviors during the dissemination process to ensure that the overall authenticity of the information is verified. Finally, by comparing with the credibility threshold, it is determined whether the information is false, further optimizing the information management process and improving the information processing efficiency of the platform.

[0198] Step 6: Automatically execute information processing measures according to the false information judgment result, including deletion, restricting dissemination, or warning the publisher; for suspected false information, trigger manual review to further confirm the authenticity of the information.

[0199] After the false information judgment is completed, the system executes corresponding information processing measures according to the judgment result (P_Final) through an automated mechanism, such as deleting information, restricting the dissemination range, or warning the publisher. This step ensures that the platform can quickly respond to false information and take timely measures to prevent the further spread of false information on the platform and affect user decisions. Automated processing not only improves processing efficiency but also reduces the cost of manual intervention, ensuring the high efficiency and real-time nature of platform management.

[0200] A platform for a public information publishing method, including:

[0201] An information collection module, which is used to obtain the information to be evaluated and publisher data, collect the historical behavior records, interaction data of the information publisher, and the information source category score; interaction data such as comments, likes, forwards, shares, etc. helps evaluate the interaction quality between the publisher and the audience, and the information source category score is used to calibrate whether the publisher is a recognized reliable source (such as government departments, well-known media, social platform users, etc.);

[0202] A content analysis module, which is used to perform sentiment analysis on the information to be evaluated, analyze the emotional bias (such as extreme words or biases) and semantic contradictions (such as self-contradictory statements) in the text; extract and check keywords in the text content to check whether it contains common keywords of false information;

[0203] External authoritative data source comparison module, which is used to obtain external authoritative data sources (such as government websites, well-known media, academic institutions, etc.), and use text similarity algorithms to compare the information content, and check the consistency between the content and the external authoritative data sources; use text similarity algorithms (such as cosine similarity, Jaccard similarity, etc.) to compare the information content, and check the consistency between the information content and the external authoritative data sources. Use the time consistency test algorithm to ensure that the release time of the information is consistent with the time of the event occurrence;

[0204] Propagation path analysis module, which is used to use social network analysis technology to construct the propagation network of information, analyze the propagation path and propagation speed of information, and detect whether there are abnormal information propagation patterns; identify abnormal nodes (such as robot accounts, malicious users, etc.) in information propagation and the centralization degree of information propagation. Analyze user behavior data to detect whether there are abnormal behaviors during the propagation process, such as batch forwarding, malicious comments, etc.;

[0205] Comprehensive credibility scoring module, which is used to comprehensively consider the credibility score P of the information source A 、the credibility score P of the content B 、the consistency reliability index P c and the propagation path reliability index P D , and use a comprehensive credibility scoring model to comprehensively evaluate the credibility of the information; assign weights to each item of data, perform weighted summation, and calculate the comprehensive credibility score. If the comprehensive credibility score is lower than the set credibility threshold, the information is determined to be false information;

[0206] Information processing module, which executes corresponding information processing measures according to the false information determination result output by the comprehensive credibility scoring module. If the information is determined to be false, the system can automatically delete the information, restrict its propagation or issue a warning notice to the publisher. For mildly false information, the platform can choose to mark the information or set access restrictions; for severely false information, measures such as deletion and banning the publisher are taken, and manual intervention is reduced through an automated processing process to ensure that the information can be processed in a timely manner. The processing process can be flexibly adjusted for different types of false information.

[0207] The above are only the preferred embodiments of the present invention, which are illustrative rather than restrictive to the present invention. Those skilled in the art understand that many changes, modifications, and even equivalents can be made within the spirit and scope defined by the claims of the present invention, but all will fall within the protection scope of the present invention.

Claims

1. A public information release method, characterized in that, It includes the following steps: Step 1: Obtain the information to be evaluated and the publisher's data, and analyze and generate the credibility score P of the information source A ; Step 2: Use sentiment analysis and keyword checking for the information to be evaluated, identify the sentiment deviation and semantic contradiction in the information to be evaluated, and analyze and generate the credibility score P of the content B ; Step 3: Obtain an external authoritative data source, use the text similarity algorithm and the time consistency test algorithm for comparison, judge the consistency between the information to be evaluated and the external authoritative data source, and output the consistency reliability index P c ; Step 4: Analyze the information dissemination path, detect the information dissemination pattern and abnormal user behavior using social network analysis technology, and output the reliability index P of the dissemination path D ; Step 5: Based on the credibility score P of the information source A , the credibility score P of the content B , the consistency reliability index P c and the reliability index P of the dissemination path D , use a comprehensive credibility scoring model to comprehensively evaluate the credibility of the information to be evaluated, obtain a comprehensive credibility score, and if the comprehensive credibility score is lower than the set credibility threshold, it is determined as false information; Step 6: According to the judgment result, take handling measures for the information to be evaluated.

2. The public information release method according to claim 1, characterized in that, The credibility score P of the information source A The generation steps include: S101: Collect data of the information to be evaluated: Obtain the specific content of the information to be evaluated and the publisher data from the information release platform. The publisher data includes the historical behavior data and interaction data of the publisher. S102: Selection and Quantification of Core Parameters: Select the core parameters of the publisher as the basis for calculating the credibility score P of the information source. The core parameters of the publisher include: A ​ Publisher's historical trustworthiness P1: Based on the verification results of the information published by the publisher in the past, calculate its credibility. The specific formula is as follows: where i is the sequential number of historical information, V i is the credibility score of each piece of historical information, and n is the number of historical information; Information source category score P2: Assign weights according to the information source type, and obtain the information source category score of each information source type through historical analysis. Interaction feedback index P3: Generate the interaction feedback index P3 by analyzing the interaction feedback of users after the information to be evaluated is published. The specific formula is as follows: Among them, C is the number of comments, L is the number of likes, S is the number of shares, and E is the entropy of the interaction type distribution. The entropy E of the interaction type distribution is: E = -(P C lnP C + P L lnP L + P S lnP S ) Among them, are respectively the proportions of the number of comments, the number of likes, and the number of shares in the total number of interactions; T is the total number of interactions, and it is: T = C + L + S. S103: Build a credibility scoring model for information sources: Based on the historical trustworthiness P1 of the publisher, the information source category score P2, and the interaction feedback index P3 obtained in S102, calculate the credibility score P of the information source through the following formula A as follows: P A = w1·P1 + w2·P2 + w3·P3 Among them, w1, w2, and w3 are the weight coefficients of the publisher's historical trustworthiness P1, information source category score P2, and interaction feedback index P3.

3. The public information release method according to claim 1, characterized in that The credibility score P of the described content B The generation steps include: S201: Conduct sentiment analysis on the content in the information to be evaluated, specifically as follows: The sentiment analysis of the information to be evaluated outputs the sentiment category and sentiment intensity; and by comparing the sentiment distribution of the content of the information to be evaluated with the normal pattern, detect whether there is a sentiment deviation; and give the corresponding sentiment analysis score P4. Its calculation formula is: P4 = sign(H)·|H| Among them, H is the sentiment intensity, and sign(H) is the polarity of the sentiment. S202: Extract the key information in the information to be evaluated, identify the negative keywords that appear in the information to be evaluated, and mark them as potential false information. The calculation formula of the keyword test score P5 is: Among them, K q is the q-th keyword, m is the total number of all keywords, Weight(K q ) is the weight of the keyword appearing in the false information, and Frequency(K q ) is the frequency of the keyword appearing in the text; S203: If semantic contradictions are detected during the analysis process, mark them as semantically inconsistent and reduce the semantic consistency score. The calculation formula of the semantic consistency score P6 is: Among them, Conflict-Score is the intensity of the semantic contradiction, and N is the number of contradictions that appear in the information to be evaluated. S204: Calculate the credibility score P of the content B , the credibility score P of the content B The calculation formula is as follows: P B = w4·P4 + w5·P5 + w6·P6 Among them, w4, w5, and w6 are the weight coefficients of the sentiment analysis score P4, keyword test score P5, and semantic consistency score P6.

4. The public information publishing method according to claim 1, wherein The specific steps of Step 3 are: S301: Obtain data related to the content of the information to be evaluated from multiple external authoritative data sources. S302: The text similarity algorithm is used to compare the information to be evaluated with the information from external authoritative data sources, and the text similarity algorithm is used to measure the content consistency of the information to be evaluated and the information from external authoritative data sources. The specific steps are as follows: First, preprocess the information to be evaluated and the external authoritative data sources. Subsequently, use the cosine similarity CS to compare the information to be evaluated with the information from external authoritative data sources. Assume that the text of the information to be evaluated is T1 and the text of the external authoritative data source is T2. Then the formula for the text similarity CS(T1, T2) is: Among them, T1,j and T2,j are the weights or word frequencies of the j-th word in the texts T1 and T2 respectively, and h is the number of different features in the texts of the information to be evaluated and the external authoritative data sources. S303: For the information to be evaluated that contains time information, adopt the time consistency checking algorithm to ensure that the time of information release is consistent with the time of event occurrence, including the following processes: Extract the key time data from the information to be evaluated; Compare the time data of the information to be evaluated with the corresponding event occurrence time in an external authoritative data source, and the probability P of the authenticity of the time data of the information to be evaluated time is as follows: where, ΔT = |T eval - T ref | is the time difference between the information to be evaluated and the external authoritative data source, and T eval is the release time of the information to be evaluated, and T ref is the occurrence time of the corresponding event of the external authoritative data source; μ is the mean of the time difference between the information to be evaluated and the external authoritative data source, σ is the standard deviation of the time difference, and T t is the threshold of the time difference; S304: Generate a consistency reliability index P by synthesizing the text similarity CS(T1, T2) and the probability P of the authenticity of the time data time , generating a consistency reliability index P c : P c = CS(T1, T2)·P time If the consistency reliability index P c meets the requirements of the consistency reliability index threshold, the information to be evaluated is considered consistent with the external authoritative data source and has a high credibility; if there is a large inconsistency, the information to be evaluated is false information or needs further verification.

5. The public information release method according to claim 4, characterized in that The method for setting the threshold of the consistency reliability index is as follows: Set the consistency reliability index threshold T based on the mean μ of the time difference between the information to be evaluated and the external authoritative data source and the standard deviation σ of the time difference, both of which are based on the normal distribution. threshold : T threshold = μ + k·σ where k is the misjudgment constant.

6. The public information publishing method according to claim 4, characterized in that The said step four includes: S401: Obtain the propagation data related to the information to be evaluated from the social network platform and construct a propagation network through user interaction data; S402: Detect the propagation mode: Based on the constructed propagation network, adopt the social network analysis algorithm to detect the propagation mode of the information, calculate the propagation centrality of each node, and identify the key nodes in the information propagation; The calculation formula for the propagation centrality of a node is: where C(v) is the propagation centrality of node v, d(u, v) is the propagation distance from node u to node v, u ∈ V is the set of nodes where node u belongs to the entire social network, and V is the set of all nodes in the social network; S403: Detect abnormal user behavior: User behavior analysis: Based on the interaction data of users on the social platform, adopt behavior analysis technology to detect whether there are abnormal user behaviors. The specific calculation formula for the abnormal behavior score BAS(v) is: Among them, B s is the behavior characteristic of the s-th user, is the average value of user behaviors, N is the total number of users. If the score exceeds the set abnormal behavior scoring threshold, the behavior is determined to be abnormal; Influence analysis in the social network: Through influence analysis, for the key influencers in the propagation of the information to be evaluated, the calculation formula for the node influence I(v) is: where N (v) is the set of neighbor nodes of node v, d(u, v) represents the propagation distance between nodes u and v, and S(u) is the propagation ability of node u; S404: Propagation path reliability index P D The calculation formula is as follows: P D = GNN[C(v), BBAS(v), I(v)] · Self - Attention[P time , T eval , T ref , ΔT] where GNN is a graph neural network used to learn the propagation path features between nodes; Self-Attention is a self-attention mechanism used to handle temporal consistency checks, evaluate the consistency between temporal differences and external data sources, and combine with P time to perform information reliability correction.

7. The public information publishing method according to claim 6, wherein, The abnormal behavior score threshold is dynamically adjusted according to the real-time data change of user behavior. The calculation formula for the abnormal behavior score threshold is: T abnormal B(t) = T abnormal B(t - 1)+α·ΔB(t) Among them, T abnormal (t) is the threshold at a moment, ΔB(t) is the increment according to the change of user behavior, and α is the learning rate.

8. The public information publishing method according to claim 1, characterized in that, The steps for the comprehensive credibility scoring model to comprehensively evaluate the information credibility are: S501: Establish a comprehensive credibility scoring model and calculate the comprehensive credibility score P 终 : P 终 = w7·P A + w8·P B + w9·P C + w 10 ·P D Among them, w7, w8, w9, w 10 are the weight coefficients of P A , P B , P C , P D respectively; S502: Set the credibility threshold. When the comprehensive credibility score is higher than the credibility threshold, the information is considered credible; when the comprehensive credibility score is lower than the credibility threshold, the information may be considered false information; S503: The result of the comprehensive credibility score will provide a basis for subsequent false information handling. Corresponding restrictive measures will be taken for false information, and credible information will continue to be released.

9. The public information release method according to claim 8, wherein Dynamic Optimization of Comprehensive Credibility Score: By continuously collecting feedback from platform users, information dissemination effects, and recognition results, regularly optimize the comprehensive credibility scoring model, and adjust the weight coefficients of P A , P B , P C , P D through machine learning and data mining techniques.

10. A platform for the public information release method according to any one of claims 1-9, characterized in that, Including: An information collection module, which is used to obtain the information to be evaluated and the publisher data, and collect the historical behavior records, interaction data of the information publisher, and the information source category score; A content analysis module, which is used to perform sentiment analysis on the information to be evaluated, analyze the sentiment deviation and semantic contradiction in the text; An external authoritative data source comparison module, which is used to obtain external authoritative data sources and use the text similarity algorithm to compare the information content to check the consistency of the content with the external authoritative data sources; A propagation path analysis module, which is used to utilize social network analysis technology to construct the propagation network of the information, analyze the propagation path and propagation speed of the information, and detect whether there are abnormal information propagation modes; Comprehensive credibility scoring module, which is used to comprehensively score the credibility P of information sources A , the credibility scoring P of the content B , the consistency reliability index P c and the propagation path reliability index P D , and use the comprehensive credibility scoring model to comprehensively evaluate the credibility of information; An information processing module, which executes corresponding information processing measures according to the false information determination result output by the comprehensive credibility scoring module.

Images