Search

CN-121092809-B - Network news false information propagation screening method

CN121092809BCN 121092809 BCN121092809 BCN 121092809BCN-121092809-B

Abstract

The invention belongs to the technical field of false information propagation screening, and particularly discloses a network news false information propagation screening method, which comprises the steps of screening key evaluation dimensions and determining weights based on historical data, generating evaluation rules, determining a conventional or emergency monitoring mode according to the amplification of forwarding quantity of information to be screened and the number of matched words of a preset keyword library, identifying an initial release source through a propagation path, evaluating source authority, content consistency and propagation abnormality in the conventional mode and outputting comprehensive credibility, and rapidly evaluating and outputting emergency credibility in an oriented mode in an emergency mode, and finally matching the credibility with an early warning interval to obtain an early warning grade; according to the invention, the source authority is evaluated based on the source information, the content consistency is judged by the text content, the transmission abnormality is evaluated by the transmission data, and the comprehensive credibility is output by combining the evaluation dimension rule, so that the multidimensional collaborative evaluation of the source, the content and the transmission is realized.

Inventors

  • PEI XIAOFEI
  • CHEN XIANGYU

Assignees

  • 南京林业大学

Dates

Publication Date
20260508
Application Date
20250815

Claims (8)

  1. 1. A network news false information propagation screening method is characterized by comprising the following steps: s1, screening key evaluation dimensions based on historical monitoring data, determining weight ratio of each evaluation dimension, and generating evaluation dimension rules according to the weight ratio; s2, obtaining the forwarding amount amplification and text content of the information to be screened in a preset period, and determining a monitoring mode according to whether the number of matched words of the forwarding amount amplification and text content and a preset keyword library exceeds a preset threshold value; S3, constructing a propagation map based on a propagation path of information to be screened to identify an initial release source; S4, under a conventional monitoring mode, respectively evaluating source authority and content consistency based on the information to be screened, the source information of the initial release source and the text content, carrying out propagation abnormality evaluation through propagation data of the information to be screened, and carrying out weighted fusion calculation on scores of all evaluation dimensions and corresponding weight ratios according to weight ratios of all evaluation dimensions to obtain comprehensive credibility; S5, under an emergency monitoring mode, carrying out directional rapid evaluation on the information to be screened based on an evaluation dimension rule, and outputting emergency credibility; s6, matching the comprehensive credibility or the emergency credibility with the credibility interval corresponding to each early warning level to obtain an early warning level of the information to be screened; The determining a monitoring mode includes: Calculating the number of matched words of the text content of the information to be screened and a preset keyword library, comparing the number of matched words with a preset matched word number threshold value, and simultaneously comparing the forwarding amount amplification of the information to be screened in a preset period with a preset forwarding amount amplification threshold value; If the forwarding quantity amplification exceeds a preset forwarding quantity amplification threshold value and the number of the matched words is larger than a preset matched word number threshold value, judging the emergency monitoring mode, otherwise judging the conventional monitoring mode; the outputting the emergency credibility includes: Respectively carrying out quantitative evaluation on the key evaluation dimensions according to the key evaluation dimensions in the evaluation dimension rule to obtain scores of the key evaluation dimensions; And obtaining the weight duty ratio of the key evaluation dimension from the weight duty ratio of each evaluation dimension, and carrying out weighted fusion calculation on the score of each key evaluation dimension and the corresponding weight duty ratio to obtain the emergency credibility.
  2. 2. The network news false information propagation screening method according to claim 1, wherein the screening key evaluation dimension comprises: Obtaining false information final judgment results of each history propagation information from the history monitoring data and false information judgment results of each evaluation dimension; Comparing the false information judgment result with the corresponding false information final judgment result; Counting the consistency times and the total times of false information judgment results of all evaluation dimensions, taking the ratio of the false information judgment results and the total times as the evaluation consistency degree of all the evaluation dimensions, and screening the evaluation dimensions with the evaluation consistency degree larger than a preset evaluation consistency degree threshold as candidate evaluation dimensions; Counting the number of candidate evaluation dimensions, and triggering secondary screening of key evaluation dimensions if the number of candidate evaluation dimensions is 0; If the number of candidate evaluation dimensions is 3, sorting the evaluation consistency of each evaluation dimension from large to small, and selecting the evaluation dimension of the first two bits of sorting as a key evaluation dimension; And for the number of other candidate evaluation dimensions, taking the evaluation dimension with the evaluation consistency degree larger than a preset evaluation consistency degree threshold value as a key evaluation dimension.
  3. 3. The network news false information propagation screening method according to claim 2, wherein the triggering key evaluation dimension secondary screening comprises the following steps: combining the evaluation dimensions in pairs, further obtaining false information judgment results of the evaluation dimension groups from the historical monitoring data, and comparing the false information judgment results with corresponding false information final judgment results; Counting the consistency times of false information judgment results of each evaluation dimension group, and taking the ratio of the false information judgment results to the total times as the evaluation consistency degree of each evaluation dimension group; Comparing the evaluation consistency of the evaluation dimension group with a preset evaluation consistency threshold, and if the evaluation dimension group with the evaluation consistency greater than the preset evaluation consistency threshold exists, sequencing the evaluation dimension group with the evaluation dimension greater than the preset evaluation consistency threshold from large to small, and further selecting the evaluation dimension in the first evaluation dimension group as a key evaluation dimension; If no evaluation dimension group with the evaluation consistency degree larger than the preset evaluation consistency degree threshold exists, sorting the evaluation consistency degree of each evaluation dimension from large to small, and selecting the evaluation dimension with the first sorting as the key evaluation dimension.
  4. 4. The method for screening false information propagation of network news according to claim 2, wherein the step of determining the weight ratio of each evaluation dimension comprises the steps of: summing the evaluation consistency of all the evaluation dimensions to obtain the total evaluation consistency; and calculating the ratio of the evaluation consistency of each evaluation dimension to the total evaluation consistency, and taking the ratio as the weight ratio of the corresponding evaluation dimension, wherein the sum of the weight ratios of the evaluation dimensions is 1.
  5. 5. The method for screening false information propagation of network news according to claim 1, wherein the identifying the initial distribution source comprises: collecting node information of information to be screened in a propagation process, taking a released account as a node based on a forwarding relation of each account in the node information, taking forwarding behavior as a directed edge, forming a propagation path network from a source to a subsequent propagation node, and marking release time of each node; And tracing back the candidate initial node with the earliest release time in the propagation path network, verifying whether the candidate initial node is a non-prepositive node of a propagation link and is consistent with the sequential logic of a downstream forwarding node through a forwarding relation, so that the candidate initial node accords with the initial starting point which is a non-prepositive node and has no contradictory condition of the sequential, and taking the candidate initial node as an initial release source.
  6. 6. The method for screening false information propagation of network news according to claim 1, wherein the step of evaluating source authority comprises the steps of: obtaining the mechanism type of the initial release source from the source information, and matching the mechanism type with the mechanism type corresponding to each type of weight to obtain the type weight of the initial release source; Acquiring source key elements of information to be screened from the source information; checking the integrity state of the source key elements, and calculating a source annotation integrity score according to the integrity degree duty ratio of the source key elements; and carrying out product calculation on the type weight of the initial release source and the source annotation integrity score to generate a source authority score.
  7. 7. The network news false information propagation screening method according to claim 1, wherein the content consistency comprises: Performing word segmentation processing on the information to be screened and the text content of the initial release source, generating text feature vectors based on a word frequency-inverse document frequency algorithm, and calculating cosine similarity values among the feature vectors to be used as overall semantic matching degree; Core keywords are extracted from the information to be screened and the text content of the initial release source respectively, the core keywords of the information to be screened and the text content of the initial release source are compared, and the key information variation type is identified by combining a preset tampering rule base; matching the mutation type with the mutation type corresponding to each mutation score, and accumulating to generate key information mutation degree based on the identified mutation type matching corresponding score; and generating a content consistency score through weighted fusion calculation based on the cosine similarity value and the key information variation degree.
  8. 8. The method for screening false information propagation of network news according to claim 1, wherein the step of evaluating propagation abnormality comprises the steps of: extracting the amplification of the forwarding amount of the information to be screened in each monitoring time period from the propagation data, extracting the maximum forwarding amount amplification from the amplification, and calculating the ratio of the maximum forwarding amount amplification to a preset forwarding amount amplification threshold value to obtain the propagation rate mutation rate; Traversing all nodes in the propagation path network, identifying the number of abnormal nodes and the total number of nodes, and taking the ratio of the abnormal nodes and the total number of nodes as propagation anomaly; and (3) carrying out product calculation on the propagation rate mutation rate and the propagation abnormality degree, and carrying out normalization processing on a calculation result to generate a propagation abnormality score.

Description

Network news false information propagation screening method Technical Field The invention belongs to the technical field of false information propagation screening, and relates to a network news false information propagation screening method. Background Web news spurious information refers to news content propagated through a web platform that does not conform to objective facts or is deliberately distorted, typical manifestations of which include, but are not limited to, spurious statements, misleading expressions, spurious data, malicious tampering with core facts, and the like. The inundation of false information of network news can mislead public cognition, disturb information spreading order, and possibly cause social panic, destroy public trust, even induce group events or economic loss, so that the false information of network news needs to be spread and screened. The method is characterized in that the text content, the corresponding comment information and the user information of news to be detected are obtained, the coding module is utilized to respectively extract the characteristics of the text content and the user comment information, and then the detection result is output through the joint attention module, so that the detection accuracy is improved through multi-source data fusion. The prior art has the following defects that 1, the current analysis of static characteristics such as text content, user comment and user information of news to be detected is mainly focused, dynamic characteristics of news in the network propagation process are not considered, typical risks of false information influenced by rapid diffusion and abnormal propagation path expansion are difficult to capture, and the comprehensive defect of detection is caused. 2. The conventional propagation and the emergency propagation are not distinguished by the current detection logic, the differential screening process cannot be started for different risk levels, meanwhile, the key evaluation dimension and the dynamic weight ratio setting are not screened based on the historical detection data, so that the effectiveness of the evaluation dimension is insufficient, the directional rapid evaluation is difficult to realize according to the emergency degree of information propagation, and the screening efficiency and the risk response accuracy are insufficient. Disclosure of Invention In view of this, in order to solve the problems set forth in the above background art, a network news false information propagation screening method is now proposed. The invention provides a network news false information propagation screening method, which comprises the following steps of S1, screening key evaluation dimensions based on historical monitoring data, determining weight duty ratio of each evaluation dimension, and generating evaluation dimension rules according to the weight duty ratio. S2, obtaining the forwarding amount amplification and text content of the information to be screened in a preset period, and determining a monitoring mode according to whether the number of matching words of the forwarding amount amplification and text content and a preset keyword library exceeds a preset threshold value. S3, constructing a propagation map based on a propagation path of the information to be screened to identify an initial release source. And S4, under a conventional monitoring mode, respectively evaluating source authority and content consistency based on the information to be screened, the source information of the initial release source and the text content, evaluating propagation abnormality through propagation data of the information to be screened, and outputting comprehensive credibility in combination with an evaluation dimension rule. And S5, under an emergency monitoring mode, carrying out directional rapid evaluation on the information to be screened based on the evaluation dimension rule, and outputting emergency credibility. And S6, matching the comprehensive credibility or the emergency credibility with the credibility interval corresponding to each early warning level to obtain the early warning level of the information to be screened. Compared with the prior art, the method has the beneficial effects that (1) the key evaluation dimension is screened based on the historical monitoring data, the effectiveness and pertinence of the evaluation dimension are improved, meanwhile, the weight duty ratio is determined through normalization processing of the evaluation consistency, the dynamic adaptation and data driving of weight distribution are realized, the continuous optimization of an evaluation system along with the updating of the historical monitoring data is ensured, and the adaptability to complex propagation scenes is further enhanced. (2) The invention triggers the emergency mode by the double conditions of the amplification super-threshold of the forwarding quantity and the super-threshold of the keyword matching num