CN113946680B - Online network rumor identification method based on graph embedding and information flow analysis - Google Patents
Online network rumor identification method based on graph embedding and information flow analysis Download PDFInfo
- Publication number
- CN113946680B CN113946680B CN202111219029.5A CN202111219029A CN113946680B CN 113946680 B CN113946680 B CN 113946680B CN 202111219029 A CN202111219029 A CN 202111219029A CN 113946680 B CN113946680 B CN 113946680B
- Authority
- CN
- China
- Prior art keywords
- network
- public opinion
- information
- graph
- time sequence
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 38
- 238000005206 flow analysis Methods 0.000 title claims abstract description 12
- 238000013528 artificial neural network Methods 0.000 claims abstract description 20
- 238000013507 mapping Methods 0.000 claims abstract description 13
- 230000007246 mechanism Effects 0.000 claims abstract description 10
- 238000007781 pre-processing Methods 0.000 claims abstract description 5
- 239000010410 layer Substances 0.000 claims description 19
- 239000011159 matrix material Substances 0.000 claims description 16
- 238000004364 calculation method Methods 0.000 claims description 13
- 239000002356 single layer Substances 0.000 claims description 4
- 230000004913 activation Effects 0.000 claims description 3
- 238000005457 optimization Methods 0.000 claims description 3
- 230000008569 process Effects 0.000 claims description 3
- 238000012545 processing Methods 0.000 claims description 3
- 230000011218 segmentation Effects 0.000 claims description 3
- 238000013527 convolutional neural network Methods 0.000 claims description 2
- 230000009193 crawling Effects 0.000 claims description 2
- 230000004927 fusion Effects 0.000 claims description 2
- 239000012633 leachable Substances 0.000 claims description 2
- 238000005096 rolling process Methods 0.000 claims description 2
- 230000002457 bidirectional effect Effects 0.000 abstract description 6
- 230000006870 function Effects 0.000 description 11
- 230000004044 response Effects 0.000 description 4
- 230000001364 causal effect Effects 0.000 description 3
- 230000009466 transformation Effects 0.000 description 3
- 230000000694 effects Effects 0.000 description 2
- 230000008451 emotion Effects 0.000 description 2
- 238000010801 machine learning Methods 0.000 description 2
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 230000000052 comparative effect Effects 0.000 description 1
- 230000002860 competitive effect Effects 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000009792 diffusion process Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000004880 explosion Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 238000011478 gradient descent method Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012163 sequencing technique Methods 0.000 description 1
- 238000012549 training Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/35—Clustering; Classification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/20—Natural language analysis
- G06F40/279—Recognition of textual entities
- G06F40/289—Phrasal analysis, e.g. finite state techniques or chunking
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/01—Social networking
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Business, Economics & Management (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Human Resources & Organizations (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Databases & Information Systems (AREA)
- Computing Systems (AREA)
- Artificial Intelligence (AREA)
- Economics (AREA)
- Computational Linguistics (AREA)
- Marketing (AREA)
- Primary Health Care (AREA)
- Strategic Management (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses an online network rumor identification method based on graph embedding and information flow analysis, which has the technical scheme that: firstly, carrying out deep preprocessing on online network public opinion events, shaving interference data, reserving characteristic information, and mapping the characteristic information to a high-dimensional space with word sense relevance; then, based on a graph embedding method, extracting public opinion network structural features contained in the hierarchical relation graph by using a deep graph convolution neural network; then establishing a bidirectional portal circulation neural network with a depth structure, and analyzing bidirectional time sequence characteristics in the public opinion information stream; and finally, fusing the network structure characteristics and the bidirectional time sequence characteristics of the public opinion by using an attention mechanism, thereby improving the accuracy of false information identification in the online network public opinion information.
Description
Technical Field
The invention belongs to the technical field of deep learning and rumor identification, and particularly relates to an online network rumor identification method based on graph embedding and information flow analysis.
Background
The online social network provides great convenience for information diffusion and builds a virtual bridge for the accelerated propagation of rumors. The rumor transmission causes great economic loss on one hand, threatens the harmonious and stable operation of society on the other hand, and becomes a social phenomenon which has to be paid attention to enough. Rumors generally refer to information which is not based on facts or is falsified deliberately, and traditional rumor identification adopts a manual method, but in the present day of information explosion, attempts to identify a complex online public opinion through manpower have almost become an impossible task. In recent years, with the push of information technology development, academia and industry began to shift the perspective of rumor identification to automated machine learning methods. One very intuitive method of identification is to analyze the public opinion information itself, which focuses only on the syntactic and semantic content carried by the public opinion information, and to extract high-dimensional linguistic features by means of modern machine learning algorithms. However, such feature-based methods are difficult to extract enough features due to the limitations of the social platform such as microblog, twitter, etc. on text length, thereby limiting the accuracy of rumor identification. In order to overcome the difficulty in feature extraction, some scholars begin to add the follow-up information such as comments, forwarding and the like of the original public opinion information to the task of public opinion identification. In the method, a tree structure of information propagation is established more classical, and then a backtracking method is adopted to gradually backtrack from leaf nodes to root nodes of information sources. However, such algorithms are not very competitive for early rumor discrimination, limited by the imperfections of the propagation tree in the early stages of public opinion propagation.
Disclosure of Invention
In order to overcome the defects of the method, the invention provides an online network rumor identification method based on graph embedding and information flow analysis, which is characterized in that the graph embedding of a propagation space structure is added while the time sequence characteristics of public opinion information are considered, and then the graph embedding and the graph embedding are fused by using an attention mechanism, so that a dynamic weighted high-dimensional public opinion characteristic representation system with the characteristic of data driving is formed, and the identification accuracy of online network rumors is improved.
In order to achieve the above purpose and achieve the identification of online network rumors, the invention is realized by the following technical scheme:
the invention is defined as follows:
definition 1 public opinion information source
Public opinion information sources are raw interpretations of information about a particular event published by individuals on a network. Formally, public opinion information sources appear at earliest on an individual or group's information publication page, using s j Representing public opinion information source, subscript j refers toSubstituting for specific public opinion events.
Definition 2 reaction information
The response information is the response of people to the public opinion information source, and is inevitably later than the public opinion information source in release time and is usedThe superscript n and the subscript j refer to the ranking of the response information at the release time and the specific public opinion event, respectively.
Definition 3 public opinion information stream
The public opinion information stream is an information set composed of all public opinion information in the time from the public opinion information source to the last response information release. In a defined information flowIn (|p| represents the amount of information), the information is ordered according to the release time. Thus, all the information flows form the whole public opinion to-be-detected set T= { T about a plurality of events 1 ,T 2 ,…,T N N represents all events.
Definition 4 public opinion hierarchical relationship diagram
The public opinion hierarchical relationship graph is a hierarchical network formed on a topological space according to the causal relationship of public opinion information release. Using V j And E is j Representing edges and points in the network, then the public opinion hierarchical relationship graph with self-connecting rings can be represented as G j ＝{V j ,E j ；φ j ,ψ j [ phi ] j Representing causal relationship between public opinion information j Representing a self-linking ring. Thus, all public opinion graphs can be expressed as g= { G 1 ,G 2 ,…,G N N represents all events.
An online network rumor identification method based on graph embedding and information flow analysis is characterized in that: firstly, carrying out deep preprocessing on online network public opinion events, shaving interference data, reserving characteristic information, and mapping the characteristic information to a high-dimensional space with word sense relevance; then, based on a graph embedding method, extracting public opinion network structural features contained in the hierarchical relation graph by using a deep graph convolution neural network; then establishing a bidirectional portal circulation neural network with a depth structure, and analyzing bidirectional time sequence characteristics in the public opinion information stream; and finally, fusing the network structure characteristics and the bidirectional time sequence characteristics of the public opinion by using an attention mechanism, thereby improving the accuracy of false information identification in the online network public opinion information.
The invention is realized by the following specific scheme:
an online network rumor identification method based on graph embedding and information flow analysis mainly comprises the following steps:
step 1) determining online network public opinion data sources, crawling structured public opinion data, and preprocessing the obtained public opinion data;
step1, deleting invalid information such as blank tweets, nonsensical comments, topic labels and the like;
step2, mapping all expression symbols in the public opinion information into literal expression according to the expression symbol release of the network standard;
step3, performing word segmentation processing on the public opinion information by utilizing the Jieba and the space, and deleting stop words at the same time;
step4, based on an SOTA algorithm, mapping the Word information to a high-dimensional space with spatial relevance by using a Word Embedding method.
Step 2) analyzing the spatial structural characteristics of online network public opinion data, abstracting the release source of original information into a central node of a network, abstracting subsequent comment, forwarding and reply information into relation nodes on the network, establishing a hierarchical structure among nodes in the network based on the logic relation of information release, mapping the established hierarchical network into an undirected graph structure with a self-connecting ring, and extracting the structural characteristics of the public opinion propagation network by means of a graph rolling neural network GCN by a graph embedding method;
step1, mapping the constructed hierarchical network into a network with self-connecting ring I N Undirected graph G of (2) j Representing the adjacency matrix of the original hierarchical network with A%, calculating an undirected graph G with self-connecting loops j The calculation method of the adjacency matrix A of (2) is as follows:
A＝A％+I N
step2 calculating the undirected graph G j The calculation method of the degree distribution matrix D comprises the following steps:
D＝∑ k A k
step3, performing graph convolution operation by utilizing the concept of Fourier transformation in the graph convolution neural network GCN, and performing first-order approximation on the obtained convolution graph to obtain an approximate graph convolution value L of the single-layer network (1) The calculation method comprises the following steps:
wherein X is Word encoding value of public opinion information, W (1) Is a weight matrix in the first layer of the GCN network,delta (g) is the activation function, typically selecting the ReLU function;
step4, the limitation of the single-layer graph convolutional neural network GCN on the network space structure presentation is limited, a layer of graph convolution layer is added, convolution operation is carried out, and a convolution value L after secondary updating is carried out (2) Namely, the extracted structural characteristics of the public opinion propagation network are obtained by the calculation method that:
wherein W is (2) Is a self-learning weight matrix in the second layer of the GCN network.
Step 3) extracting the release time of each information in online network public opinion data, constructing public opinion information stream based on the release time, taking the obtained information stream as input, establishing a Bi-directional gate cycle neural network Bi-GRU with a depth structure, respectively extracting time sequence features in the public opinion information in the forward and reverse time sequence angles, and fusing the time sequence features in the two dimensions to form the angle overall feature presentation of the public opinion information stream in the time sequence;
step1, sorting the public opinion events according to the release time, and establishing a time sequence information stream which has complete description for a certain public opinion event.
Step2, inputting public opinion information into a gate-cycle neural network GRU in turn according to the forward information flow direction, and extracting time sequence characteristics h 'contained in the forward information flow' a,t Calculated by the following formula:
wherein U is z ,W z ,U r ,W r ,U h％ And W is h％ Respectively represent the leachable weights in the GRU units, e represents the element-by-element multiplication operation, h' a,t-1 Is the timing characteristic of the last point in time,is h' a,t Is the hidden layer alternative state of +.>Representing the states of the reset gate and the update gate, respectively, σ (·) is a Sigmoid function;
step3, inputting public opinion information into the gate cycle neural network GRU in turn according to the reverse information flow direction, wherein the extracted reverse information flow comprises a time sequence characteristic h' a,t The calculation method is the same as that in the forward direction;
step4, connecting the forward and reverse information flow time sequence characteristics to form the overall characteristic presentation h of the public opinion information flow on the time sequence a,t 。
Step 4) adopting an Attention mechanism, fusing graph structural features obtained by embedding graphs and information flow time sequence features of overall presentation to form a dynamic weighted high-dimensional public opinion feature representation system characterized by data driving, and taking the output of the dynamic weighted high-dimensional public opinion feature representation system as a final mixed feature presentation mode of rumor identification;
step1 establishing a sum StepThe portal circulation neural network with identical parameters in the step 3) is used for embedding the structural characteristics L of the public opinion propagation network based on the graph embedding method obtained in the step 2) (2) As input, a time series characteristic h of the sum public opinion information stream is obtained a,t Network output h of exactly the same dimension b,t ；
Step2 fusion h by using the Attention mechanism a,t And h b,t Obtaining h t The final mixed characteristic is presented, and the specific calculation method is as follows:
wherein, gamma and W h Is a learnable weight parameter in the network.
And 5) taking the fully connected network as an output layer, matching with a Softmax function to map the output to a probability space, defining a Loss function, judging a virtual space distance between a probability result predicted by the model and real data, taking minimization of the Loss function as an optimization direction of the model, and continuously adjusting all relevant weight parameters in the network.
The Loss function is:
wherein N is s Is the sample size, C is the sample classification, y c (x n ) Is the number of samples that fall into a certain class,is the number of samples that the model predicts as a particular class, |θ| 2 Is the second order norm of all the learnable parameters to avoid overfitting.
Drawings
Fig. 1 is a flow chart of an online network rumor authentication method based on graph embedding and information flow analysis.
Detailed Description
In order to make the technical route, the structural features and the effects of the present invention clear, the present invention will be further described with reference to the accompanying drawings and specific examples, so that the related art can better understand and practice the present invention, but the examples used are not limited thereto.
As shown in fig. 1, an online network rumor authentication method based on graph embedding and information flow analysis mainly comprises the following steps:
step 1), selecting two typical online network social platforms of Xinlang microblogs (Chinese) and Twitter (English), respectively obtaining published relevant network published data about series public opinion events, and deleting invalid information such as blank tweets, nonsensical comments, topic labels and the like in the data; then mapping all the emoticons in the public opinion information into literal expression according to the emotion interpretation of the emotion symbol of the network standard; then, utilizing Jieba and space to perform word segmentation processing on the public opinion information, and deleting stop words at the same time; finally, based on an SOTA algorithm, mapping Word information to a high-dimensional space with spatial correlation by using a Word Embedding method, wherein the dimension of the Word Embedding is set to be 300;
step 2) respectively analyzing the spatial structural characteristics of the obtained public opinion data on the newwave microblogs and the Twitter, abstracting the release source of the original information into a central node of the network, abstracting the subsequent comment, forwarding and reply information into relation nodes on the network, establishing a hierarchical structure among the nodes in the network based on the logic relation of information release, and mapping the constructed hierarchical network into an undirected graph structure with a self-connecting ring;
constructing an adjacency matrix according to the established self-loop undirected graph mechanism, wherein if a causal relationship of information release exists between two adjacent nodes in the graph, the value in the adjacency matrix is marked as 1, otherwise, the value is 0; because the adjacent matrix is in a sparse structure, a construction mode of the sparse matrix is used here, so that the memory consumption is saved;
establishing a degree distribution matrix according to the adjacent matrix;
constructing a graph convolution neural network to extract structural features of a public opinion propagation network, wherein a graph embedding mode is carried out by using a two-layer stacked GCN structure, wherein the Fourier transformation is approximated in a first order, the hidden layer dimension in the GCN is set to be 256, and the output layer dimension is set to be 128;
and 3) respectively sequencing the public opinion events on the New wave microblogs and the Twitter according to the release time of the public opinion events, and establishing a time sequence information stream which has complete description on a certain public opinion event. Then taking the obtained information stream as input, establishing a Bi-directional gate cycle neural network Bi-GRU with a depth structure, respectively extracting time sequence features in the public opinion information in the forward direction and the reverse direction time sequence angles, and fusing the time sequence features in the two dimensions to form overall feature presentation of the public opinion information stream in the time sequence angles;
dropout operation is added in the realization process of the gate cycle neural network GRU to prevent overfitting; considering the inconsistency of the public opinion information length, the information flow needs to be subjected to a Padding operation, so that the information in a specific public opinion event is unified to be the same length, the Padding value is set to 0, the Bi-GRU layer number is set to 2, and the hidden layer dimension is 256;
step 4) establishing a gate cycle neural network with the same parameters as those in the step 3), taking the structural characteristics of the public opinion propagation network based on the graph embedding method obtained in the step 2) as input, and carrying out dimension transformation on the structural characteristics to obtain a network output with the same time sequence characteristic dimension as that of the public opinion information flow;
adopting an Attention mechanism, fusing graph structural features obtained by embedding graphs and information flow time sequence features of overall presentation to form a dynamic weighted high-dimensional public opinion feature representation system with data driving as a characteristic, and taking the output of the dynamic weighted high-dimensional public opinion feature representation system as a final mixed feature presentation mode of rumor identification; an activation function of the Attention network adopts Tanh, the dimension of an input layer is 256 x 2, the output is Attention Score, and the dimension is 1;
and 5) taking the fully connected network as an output layer, matching with a Softmax function to map the output to a probability space, defining a Loss function, judging a virtual space distance between a probability result predicted by the model and real data, taking minimization of the Loss function as an optimization direction of the model, and continuously adjusting all relevant weight parameters in the network.
The Batch Size of the model was set to 128 and the optimizer was chosen as Adam, beta 1 And beta 2 Set to 0.9 and 0.999, respectively; the initial learning rate is set to be 2e-3 by using a random gradient descent method, and gradually descends along with the training process;
the evaluation index adopts Accuracy Accuracy, accuracy Precision, recall rate Recall and F1 score, and the specific calculation formula is as follows:
wherein TP is the number of positive cases with correct prediction, FP is the number of positive cases with incorrect prediction, TN is the number of negative cases with correct prediction, FN is the number of negative cases with incorrect prediction;
for predictive analysis of rumors in public opinion, a 4-classification subdivision scheme was used, namely, non-rumors (NR), false Rumors (FR), true Rumors (TR) and Unverified Rumors (UR), respectively;
comparative experimental results with the relevant rumor identification method are shown in Table 1 below
Table 1 rumor discrimination and comparison experiment results
Microblog data set
Twitter dataset
The rumor identification method has the advantages that the rumor identification effect is highest under each classification, and compared with a comparison method, the rumor identification accuracy in online network public opinion events is improved.
While the invention has been described with reference to certain preferred embodiments, it will be understood by those skilled in the art that the invention is not limited thereto, and that modifications may be made without departing from the spirit of the invention.
Claims (3)
1. An online network rumor identification method based on graph embedding and information flow analysis is characterized by comprising the following specific steps:
step 1) determining online network public opinion data sources, crawling structured public opinion data, and preprocessing the obtained public opinion data;
step 2) analyzing the spatial structural characteristics of online network public opinion data, abstracting the release source of original information into a central node of a network, abstracting subsequent comment, forwarding and reply information into relation nodes on the network, establishing a hierarchical structure among nodes in the network based on the logic relation of information release, mapping the established hierarchical network into an undirected graph structure with a self-connecting ring, and extracting the structural characteristics of the public opinion propagation network by means of a graph rolling neural network GCN by a graph embedding method;
step 2) extracting structural features of public opinion propagation network based on graph embedding method, and the specific method is as follows:
step1, mapping the constructed hierarchical network into a network with self-connecting ring I N Undirected graph G of (2) j ToRepresenting an adjacency matrix of an original hierarchical network, computing an undirected graph G with self-connecting loops j The calculation method of the adjacency matrix A of (2) is as follows:
step2 calculating the undirected graph G j The calculation method of the degree distribution matrix D comprises the following steps:
D＝∑ k A k
step3, performing Fourier transform in the graph roll-up neural network GCNPerforming graph convolution operation, and performing first-order approximation on the obtained convolution graph to obtain an approximate graph convolution value L of a single-layer network (1) The calculation method comprises the following steps:
wherein X is Word encoding value of public opinion information, W (1) Is a weight matrix in the first layer of the GCN network,delta (·) is the activation function, typically selecting a ReLU function;
step4, the limitation of the single-layer graph convolutional neural network GCN on the network space structure presentation is limited, a layer of graph convolution layer is added, convolution operation is carried out, and a convolution value L after secondary updating is carried out (2) Namely, the extracted structural characteristics of the public opinion propagation network are obtained by the calculation method that:
wherein W is (2) Is a self-learning weight matrix in the second layer of the GCN network;
step 3) extracting the release time of each information in online network public opinion data, constructing public opinion information stream based on the release time, taking the obtained information stream as input, establishing a Bi-directional gate cycle neural network Bi-GRU with a depth structure, respectively extracting time sequence features in the public opinion information in the forward and reverse time sequence angles, and fusing the time sequence features in the two dimensions to form overall feature presentation of the public opinion information stream in the time sequence angle;
step 3) overall characteristic presentation of public opinion information stream on time sequence, the specific method is as follows:
step1, ordering public opinion events according to release time, and establishing a time sequence information stream which has complete description for a certain public opinion event;
step2, inputting public opinion information into a gate-cycle neural network GRU in turn according to the forward information flow direction, and extracting time sequence characteristics h 'contained in the forward information flow' a,t Calculated by the following formula:
wherein U is z ,W z ,U r ,W r ,And->Respectively represent the leachable weights in the GRU units, +.' a,t-1 Is the timing characteristic of the last time point, +.>Is h' a,t Is the hidden layer alternative state of +.>Representing the states of the reset gate and the update gate, respectively, σ (·) is a Sigmoid function;
step3, according to the reverse information flow direction, sequentially mixing the public opinionThe information is input into the gate cycle neural network GRU, and the extracted reverse information flow contains a time sequence characteristic h a,t The calculation method is the same as that in the forward direction;
step4, connecting the forward and reverse information flow time sequence characteristics to form the overall characteristic presentation h of the public opinion information flow on the time sequence a,t ；
Step 4) adopting an Attention mechanism, fusing graph structural features obtained by embedding graphs and information flow time sequence features which are totally presented to form a dynamic weighted high-dimensional public opinion feature representation system which takes data driving as a characteristic, and taking the dynamic weighted high-dimensional public opinion feature representation system as a final mixed feature presentation mode for rumor identification;
and 4) adopting a mixed feature presentation mode fused by an Attention mechanism, wherein the specific method is as follows:
step1, establishing a portal circulation neural network with the same parameters as those in the Step 3), and embedding the structural characteristics L of the public opinion propagation network based on the graph embedding method obtained in the Step 2) (2) As input, a time series characteristic h of the sum public opinion information stream is obtained a,t Network output h of exactly the same dimension b,t ；
Step2 fusion h by using the Attention mechanism a,t And h b,t Obtaining h t The final mixed characteristic is presented, and the specific calculation method is as follows:
wherein, gamma and W h Is a learnable weight parameter in the network;
and 5) taking the fully connected network as an output layer, matching with a Softmax function to map the output to a probability space, defining a Loss function, judging a virtual space distance between a probability result predicted by the model and real data, taking minimization of the Loss function as an optimization direction of the model, and continuously adjusting all relevant weight parameters in the network.
2. The online network rumor authentication method based on graph embedding and information flow analysis according to claim 1, wherein the step 1) preprocessing of online network public opinion data comprises the following specific processes:
step1, deleting invalid information such as blank tweets, nonsensical comments, topic labels and the like;
step2, mapping all expression symbols in the public opinion information into literal expression according to the expression symbol release of the network standard;
step3, performing word segmentation processing on the public opinion information by utilizing the Jieba and the space, and deleting stop words at the same time;
step4, based on an SOTA algorithm, mapping the Word information to a high-dimensional space with spatial relevance by using a Word Embedding method.
3. The online network rumor authentication method based on graph embedding and information flow analysis according to claim 1, wherein the determining of the Loss function in step 5) specifically comprises:
wherein N is s Is the sample size, C is the sample classification, y c (x n ) Is the number of samples that fall into a certain class,is the number of samples that the model predicts as a particular class, |θ| 2 Is the second order norm of all the learnable parameters to avoid overfitting.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111219029.5A CN113946680B (en) | 2021-10-20 | 2021-10-20 | Online network rumor identification method based on graph embedding and information flow analysis |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111219029.5A CN113946680B (en) | 2021-10-20 | 2021-10-20 | Online network rumor identification method based on graph embedding and information flow analysis |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113946680A CN113946680A (en) | 2022-01-18 |
| CN113946680B true CN113946680B (en) | 2024-04-16 |
Family
ID=79331674
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111219029.5A Active CN113946680B (en) | 2021-10-20 | 2021-10-20 | Online network rumor identification method based on graph embedding and information flow analysis |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113946680B (en) |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109685153A (en) * | 2018-12-29 | 2019-04-26 | 武汉大学 | A kind of social networks rumour discrimination method based on characteristic aggregation |
| WO2019205318A1 (en) * | 2018-04-25 | 2019-10-31 | 平安科技（深圳）有限公司 | Public opinion information classification method and apparatus, computer device, and storage medium |
| WO2019227710A1 (en) * | 2018-05-31 | 2019-12-05 | 平安科技（深圳）有限公司 | Network public opinion analysis method and apparatus, and computer-readable storage medium |
| WO2020000847A1 (en) * | 2018-06-25 | 2020-01-02 | 中译语通科技股份有限公司 | News big data-based method and system for monitoring and analyzing risk perception index |
| CN112069397A (en) * | 2020-08-21 | 2020-12-11 | 三峡大学 | Rumor detection method combining self-attention mechanism with generation of confrontation network |
| CN112231562A (en) * | 2020-10-15 | 2021-01-15 | 北京工商大学 | Network rumor identification method and system |
| WO2021073271A1 (en) * | 2019-10-17 | 2021-04-22 | 平安科技（深圳）有限公司 | Public opinion analysis method and device, computer device and storage medium |
| CN112766359A (en) * | 2021-01-14 | 2021-05-07 | 北京工商大学 | Word double-dimensional microblog rumor recognition method for food safety public sentiment |
| CN113204641A (en) * | 2021-04-12 | 2021-08-03 | 武汉大学 | Annealing attention rumor identification method and device based on user characteristics |
-
2021
- 2021-10-20 CN CN202111219029.5A patent/CN113946680B/en active Active
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2019205318A1 (en) * | 2018-04-25 | 2019-10-31 | 平安科技（深圳）有限公司 | Public opinion information classification method and apparatus, computer device, and storage medium |
| WO2019227710A1 (en) * | 2018-05-31 | 2019-12-05 | 平安科技（深圳）有限公司 | Network public opinion analysis method and apparatus, and computer-readable storage medium |
| WO2020000847A1 (en) * | 2018-06-25 | 2020-01-02 | 中译语通科技股份有限公司 | News big data-based method and system for monitoring and analyzing risk perception index |
| CN109685153A (en) * | 2018-12-29 | 2019-04-26 | 武汉大学 | A kind of social networks rumour discrimination method based on characteristic aggregation |
| WO2021073271A1 (en) * | 2019-10-17 | 2021-04-22 | 平安科技（深圳）有限公司 | Public opinion analysis method and device, computer device and storage medium |
| CN112069397A (en) * | 2020-08-21 | 2020-12-11 | 三峡大学 | Rumor detection method combining self-attention mechanism with generation of confrontation network |
| CN112231562A (en) * | 2020-10-15 | 2021-01-15 | 北京工商大学 | Network rumor identification method and system |
| CN112766359A (en) * | 2021-01-14 | 2021-05-07 | 北京工商大学 | Word double-dimensional microblog rumor recognition method for food safety public sentiment |
| CN113204641A (en) * | 2021-04-12 | 2021-08-03 | 武汉大学 | Annealing attention rumor identification method and device based on user characteristics |
Non-Patent Citations (2)
| Title |
|---|
| 基于RNN的网络舆情谣言判断研究;王淼;郭阳明;陈泽林;钟林龙;;电脑知识与技术;20200825(24);全文 * |
| 社交网络舆情中意见领袖主题图谱构建及关系路径研究――基于网络谣言话题的分析;王晰巍;张柳;韦雅楠;王铎;;情报资料工作;20200325(02);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113946680A (en) | 2022-01-18 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Priyadarshini et al. | A novel LSTM–CNN–grid search-based deep neural network for sentiment analysis | |
| CN107608956B (en) | Reader emotion distribution prediction algorithm based on CNN-GRNN | |
| CN109558487A (en) | Document Classification Method based on the more attention networks of hierarchy | |
| Harjule et al. | Text classification on Twitter data | |
| CN111966825A (en) | Power grid equipment defect text classification method based on machine learning | |
| CN110851593A (en) | Complex value word vector construction method based on position and semantics | |
| CN111241425A (en) | POI recommendation method based on hierarchical attention mechanism | |
| CN110569355B (en) | Viewpoint target extraction and target emotion classification combined method and system based on word blocks | |
| CN113094502A (en) | Multi-granularity takeaway user comment sentiment analysis method | |
| Al Maruf et al. | Emotion detection from text and sentiment analysis of Ukraine Russia war using machine learning technique | |
| Jayakody et al. | Sentiment analysis on product reviews on twitter using Machine Learning Approaches | |
| Başarslan et al. | Sentiment analysis with ensemble and machine learning methods in multi-domain datasets | |
| Parvathi et al. | Identifying relevant text from text document using deep learning | |
| CN110889505A (en) | Cross-media comprehensive reasoning method and system for matching image-text sequences | |
| Jeyakarthic et al. | Optimal bidirectional long short term memory based sentiment analysis with sarcasm detection and classification on twitter data | |
| Saha et al. | The corporeality of infotainment on fans feedback towards sports comment employing convolutional long-short term neural network | |
| Shan | Social Network Text Sentiment Analysis Method Based on CNN‐BiGRU in Big Data Environment | |
| Zhang et al. | Predicting and visualizing consumer sentiments in online social media | |
| CN113946680B (en) | Online network rumor identification method based on graph embedding and information flow analysis | |
| CN116227486A (en) | Emotion analysis method based on retrieval and contrast learning | |
| Jasim et al. | Analyzing Social Media Sentiment: Twitter as a Case Study | |
| Lisjana et al. | Classifying complaint reports using rnn and handling imbalanced dataset | |
| CN116595125A (en) | Open domain question-answering method based on knowledge graph retrieval | |
| Tamrakar | An analytical study of feature extraction techniques for student sentiment analysis | |
| Le-Minh et al. | Aspect-based sentiment analysis using mini-window locating attention for vietnamese e-commerce reviews |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |